Mental Health in Temporary Construction Workers

Qinxue Li

A thesis

submitted in partial fulfillment of the

requirements for the degree of

Master of Science in Construction Management

University of Washington

2018

Committee:

Ken-Yu Lin

Edmund Seto

Ahmed Abdel-Aziz

Program Authorized to Offer Degree:

Construction Management

©Copyright 2018

Qinxue Li

University of Washington

Abstract

Mental Health in Temporary Construction Workers

Qinxue Li

Chair of the Supervisory Committee:

Ken-Yu Lin

Department of Construction Management

Construction industry is characterized as not only physically but also mentally demanding with

industry specific characteristics such as heavy workload, long working hours, work-family

conflicts etc. However, inadequate research has been done to address the mental health aspect of

the construction industry, where attention has been dominantly focused on physical health. On the

other hand, the variance in terms of labor demands also distinguishes the construction industry

from the rest and results in the second highest temporary employment rate based on the report

from The Center for Construction Research and Training (CPWR) in 2015. Since the temporary

construction workers are reported to have been earning less income, receiving less benefits and

experiencing higher job insecurity, this study looks into the mental health of this disadvantaged

population.

In the literature review, characteristics of the overall as well as the temporary construction

workforce are identified. The measurement of the mental health performance is determined to be

the Kessler Index (K6) and the odds of serious mental illness (SMI). Meanwhile, the variables

which are considered to be associated with mental health are also teased out from previous

studies, which consist of demographics, annual income, housing status, health status and others.

The two data sources identified are the Medical Expenditure Survey (MEPS) and Behavioral Risk

Factor Surveillance System-Washington State (BRFSS-WA). Due to the information availability

of the variables of interest, two different models are proposed, with one at the national level based

on MEPS and one at the Washington State level based on BRFSS-WA.

At the national level, the comparisons are made among the non-construction, permanent

construction and temporary construction workforce while at the Washington State level, the

comparison is made between non-construction and construction workers. Descriptive analysis is

performed on all the variables to create a general profile at the national level and the state level.

Linear regression and logistic regression are also performed in order to test the significance of the

difference at the two levels.

Based on the analysis, it is found that, at the national level, permanent construction workers have

significantly lower mean K6 score than the non-construction workers. Temporary construction

workers are found to have lower mean K6 score than non-construction workers and higher mean

K6 score than permanent construction workers, but neither of the difference is significant. As for

the odds of SMI, permanent construction workers tend to have significantly lower odds than the

non-construction workers. Temporary construction workers tend to have higher odds than the

other two groups, but neither of the difference is significant. At the Washington State level,

construction workers are found to have lower mean K6 score and odds of SMI, but neither of

them is significant. In addition, the significance of the covariates are also discussed in the

analysis.

Moreover, the lessons learnt from the data handling process are discussed and recommendations

are listed out. Limitations of this study and suggestions for future research are also presented at

the end.

Table of Contents

1 Introduction ................................................................................................................. 1

2 Literature Review ........................................................................................................ 4

2.1 Total worker health (TWH) ............................................................................................. 4

2.1.1 Characteristics of construction industry .................................................................. 4

2.1.2 Health Status of Construction Workers ................................................................... 6

2.2 Mental health ................................................................................................................... 7

2.2.1 Potential covariates .................................................................................................. 7

2.2.2 Potential pathway towards negative effect on health .............................................. 7

2.3 Mental health in temporary construction workers ........................................................... 8

3 Methodology ................................................................................................................ 9

3.1 Available data sources ..................................................................................................... 9

3.1.1 MEPS ....................................................................................................................... 9

3.1.2 BRFSS-WA ........................................................................................................... 10

3.1.3 Available information and potential gaps .............................................................. 10

3.2 Proposed models ............................................................................................................ 15

4 Analysis and Results .................................................................................................. 16

4.1 At National Level .......................................................................................................... 16

4.1.1 Data Extraction ...................................................................................................... 16

4.1.2 Descriptive Analysis .............................................................................................. 19

4.1.3 Inferential Analysis ............................................................................................... 28

4.2 At Washington State Level ............................................................................................ 34

4.2.1 Data Extraction ...................................................................................................... 34

4.2.2 Descriptive Analysis .............................................................................................. 37

4.2.3 Inferential Analysis ............................................................................................... 45

5 Reflections on the Existing Data Sets ....................................................................... 52

6 Conclusion ................................................................................................................. 55

7 Bibliography .............................................................................................................. 57

8 Appendix ................................................................................................................... 62

List of Figures

Figure 1 Distribution of all the occupations (At national level, 2008-2015) ................................. 16

Figure 2 Distribution of Non-construction vs. Construction (At national level, 2008-2015) ....... 17

Figure 3 Distribution of age (At national level, 2008-2015) ......................................................... 20

Figure 4 Distribution of education (At national level, 2008-2015) ............................................... 21

Figure 5 Distribution of annual personal wage (At national level, 2008-2015) ............................ 22

Figure 6 Non- vs. permanent vs. temporary construction workers with <25k annual personal wage

(At national level, 2008-2015) ...................................................................................................... 22

Figure 7 Distribution of annual family income (At national level, 2008-2015) ............................ 23

Figure 8 Distribution of perceived general health (At national level, 2008-2015)........................ 24

Figure 9 Distribution of current smoking status (At national level, 2008-2015) ......................... 24

Figure 10 Distribution of health insurance coverage (At national level, 2008-2015) ................... 25

Figure 11 Health insurance coverage in non- vs. permanent vs. temporary construction workers

(At national level, 2008-2015) ...................................................................................................... 25

Figure 12 Distribution of all the occupations (At WA level, 2011-2016) ..................................... 35

Figure 13 Distribution of Non-construction vs. Construction (At WA level, 2011-2016) ............ 35

Figure 14 Distribution of age (At WA level, 2011-2016) ............................................................. 37

Figure 15 Distribution of education (At WA level, 2011-2016) ................................................... 38

Figure 16 Distribution of annual family income (At WA level, 2011-2016) ................................ 39

Figure 17 Annual family income non-construction vs. construction (At WA level, 2011-2016) . 39

Figure 18 Distribution of perceived general health (At WA level, 2011-2016) ............................ 40

Figure 19 Distribution of current smoking status (At WA level, 2011-2016) ............................... 40

Figure 20 Trend of health insurance coverage (At WA level, 2011-016) ..................................... 41

Figure 21 Health insurance coverage rate in non-construction vs. construction (At WA level,

2011-2016) .................................................................................................................................... 41

Figure 22 Distribution of housing status (At WA level, 2011-2016) ............................................ 42

Figure 23 Trend of housing status in non-construction vs. construction (At WA level, 2011-2016)

....................................................................................................................................................... 42

List of Tables

Table 1 Date Availability at National Level .................................................................................. 13

Table 2 Date Availability at WA Level ......................................................................................... 14

Table 3 Mean K6 Score among Different Occupations (At national level, 2008-2015) ............... 17

Table 4 Prevalence of SMI among Different Occupations (At national level, 2008-2015) .......... 18

Table 5 Descriptive Summary of All the Variables (At national level, 2008-2015) ..................... 26

Table 6 Mean K6 Score and OR of SMI among three groups with covariates adjusted (At national

level, 2008-2015). .......................................................................................................................... 29

Table 7 Mean K6 Score among Different Occupations (At WA level, 2011-2016) ..................... 36

Table 8 Prevalence of SMI among Different Occupations (At WA level, 2011-2016) ................ 36

Table 9 Descriptive Summary of All the Variables (At WA level, 2011-2016) ........................... 43

Table 10 Mean K6 Score and OR of SMI among two groups with covariates adjusted (At WA

level, 2011-2016 ............................................................................................................................ 46

Table 11 Descriptive Summary of All the Variables (At national level, 2008) ............................ 62

Table 12 Descriptive Summary of All the Variables (At national level, 2012) ............................ 64

Table 13 Descriptive Summary of All the Variables (At national level, 2015) ............................ 66

Table 14 Descriptive Summary of All the Variables (At WA level, 2011) .................................. 68

Table 15 Descriptive Summary of All the Variables (At WA level, 2012) .................................. 70

Table 16 Descriptive Summary of All the Variables (At WA level, 2013) .................................. 73

Table 17 Descriptive Summary of All the Variables (At WA level, 2014) .................................. 76

Table 18 Descriptive Summary of All the Variables (At WA level, 2015) .................................. 79

Table 19 Descriptive Summary of All the Variables (At WA level, 2016) .................................. 82

1

1 Introduction

Total worker health is considered a comprehensive approach to improve worker safety and

health (CDC, 2016). However, majority of the research mainly focused on physical health as

the outcome of interest, and limited attention was paid to mental health, which is an essential

element of general health. A positive mental health enables us to realize our potential fully,

deal with life stress properly, work productively and contribute to our communities

meaningfully (MentalHealth.gov, 2017).

Felter et al. (2016) reviewed studies from 1990 to 2015 in order to evaluate the effectiveness

of TWH interventions and out of the 24 eligible studies identified, only six of them

considered mental health as the outcome of interest. Among the six, Maes et al. (1998),

Eriksen et al. (2002), Palumno et al. (2012), Coffeng et al., 2014, and Hammer et al. (2015)

focused on job stress whereas Olson et al. (2015) focused on depression and psychosocial

stress. With regard to the populations selected by the six studies, only Hammer et al. (2015)

enrolled primarily construction workers while the rest focused on health care and serve,

finance, and transportation workers.

On the other hand, construction workers tend to have high work demand, low job control,

poor organizational control and work-family conflicts (Todd et al., 2014). These job

conditions make the workers more prone to reduced mental health due to the high job stress

level (NIOSH, 2014). Additionally, within the construction industry, temporary workers are

further disadvantaged with lower earnings, fewer benefits, higher exposure to hazards and

lower job insecurity (CPWR, 2015), making them seemingly more vulnerable when being

compared with full-time workers.

Temporary workers, whose job lasts for only a limited amount of time or until the

completion of a project, enable construction companies to adjust the labor demand with

relatively low costs. In 2015, the U.S. construction workforce consisted of 15.5% temporary

workers, with a growth more than 40% from 2003. When being compared with the other

industries, the percentage of temporary workers employed in construction is 70% and 40%

higher in 2015 and 2003 respectively (CPWR, 2015). As the U.S. continues to build its

infrastructure, demand for the construction workforce is high and the employment of

temporary workers is on the rise. While urban cities offer plentiful employment

2

opportunities for the construction industry, they also bring forth complicated economic,

transportation, housing … etc. challenges to all who work and/or reside in the cities.

In order to better understand the influence of temporary employment on the mental health of

construction workers under the emerging U.S. social-economic environment, the mental

health performance among non-, permanent and temporary construction workers is

investigated in this thesis. Specifically, data were gathered and analyzed both for across the

nation and at the Washington State level. The 2016 GDP growth in the Washington State

was the highest in the nation, and the state’s 2015 employment and wage growth was also

the top of the nation, with most hiring in construction, information technology and a couple

of other industries.

Since many factors could be at play, adequate measurement of mental health performance

and the potential variables (or covariates, in statistical terms) are identified based on the

literature review described in Chapter 2. Knowing how employment status influences

construction workers’ mental health at the juncture of these potential variables enables

stakeholders to see how the issue could be multifaceted and cannot be addressed by a unitary

method.

The data source for the national level analysis is the Medical Expenditure Survey (MEPS),

which is a national wide survey collecting basic household and insurance information of the

civilian non-institutional population in the U.S. The data source for the Washington State

level analysis is the Behavioral Risk Factor Surveillance System-Washington State (BRFSS-

WA), which gathers information of people’s health changes in the state. Although data on

the employment status (temporary versus permanent) are not available through BREFF-WA

for the Washington State, the comparison of non-construction and construction labor force

was still conducted in the thesis to provide a point of inference on the mental health for the

temporary construction workers within the Washington State. Chapter 3 discusses the overall

research methodology, including the data sources.

Chapter 4 explains the data extraction approach, presents the data descriptively, and

elucidates how the inferential analysis was conducted statistically. Data extraction

essentially involves selecting meaningful data from the sources and preparing the selected

data for statistical analysis. The descriptive analysis summarizes the distributions of all the

3

research covariates. The inferential analysis applies linear and logistic regressions to test the

statistical significance of the temporary employment on the mental health performance with

all the covariates being adjusted.

Chapter 5 reflects upon the lessons learnt working with the existing data sets. For example,

even though a standardized occupation classification system is adopted by the sources from

which this thesis drew data, groupings of the data at the sources are quite different, making

integrated data analysis impossible if not extremely difficult. Finally in Chapter 6, the

findings are concluded along with the research limitation discussion and recommendations

for the future work.

4

2 Literature Review

2.1 Total worker health (TWH)

CDC (2016) described well-being as judging life positively and feeling good, which

was built upon social, economic and personal development enabled by a healthy

condition. Through promoting health, people gain more access to individual, social as

well as environmental resources which may ultimately improve their well-being.

NIOSH launched the TWH Program in June 2011 with a vision to not only protect

workers from hazards, but also promote their health and well-being. In this approach,

work is considered as a social determinant of health which can exert an important

impact on the well-being of workers through work-related factors such as income,

work load, stress level, interaction among the coworkers etc.

Under the scope of promoting the TWH, research has been done across all industries,

but with limited attention to address the mental health issue. Felter et al. (2016)

conducted a literature review on TWH interventions across all industries and out of

the 24 eligible studies, only six studies considered mental health as the outcome of

interest (Maes et al., 1998; Eriksen et al., 2002; Palumno et al., 2012; Coffeng et al.,

2014; Hammer et al., 2015). There were four studies which enrolled primarily

construction workers in the 24 eligible studies, but only one considered work-life

stress reduction as the outcome of interest (Hammer et al. 2015), while the rest mainly

focused on physical health- smoking status (Barbeau et al. 2006, Sorensen et al. 2007)

and shoulder pain (Borstad et al. 2009).

2.1.1 Characteristics of the construction industry

Unique characteristics of the labor force distinguish the construction industry

from the rest. In order to have a better understanding of the health conditions of

those who work in the construction industry, it is crucial to go through the overall

profile of this industry.

Aging workforce

5

The pace of aging in the construction labor force has exceeded that for the

general industries. In the sixth edition of The Construction Chart Book

(CPWR, 2018), CPWR provided a general overview of the construction

industry with data collected up to 2015. It reported that the average age of

workers increased by 6.5 years between 1985 and 2015 in the construction

industry comparing to 4.9 years in all U.S. workers. The shift in the age

distribution was also found within the construction labor force between

1985 and 2015, with 59% increase in the age group of 45 to 64 years and

67%, 49% and 32% decrease in the age groups of 16 to 19 years, 20 to 24

years, and 25 to 34 years respectively.

Health insurance coverage

The rate of health insurance coverage in construction is lower compared

with all industries combined. In 2015, 78.3% wage-and-salary workers in

construction had health insurance coverage, compared to 89.9% in the

general industries. As for the self-employed workers who constituted

24.5% of the construction workers, only 74% was covered by health

insurance (CPWR, 2018).

Family income as a percentage of poverty line

The income level in construction tends to be more likely below the poverty

line. The CPWR reported, from 2012 to 2014, that more than 7% of

construction workers lived below the federal poverty level, compared to

6.1% in all industries. Nearly 25% of the constriction workers had a

household income less than two times of the federal poverty level,

compared with about 20% in all industries (CPWR, 2016). For instance, in

2014, the federal poverty level was $11,670 with two persons in household

was $15,730 and $23,850 with, two times less than which were $3,890 and

$7950 respectively (ASPE, 2014)

Temporary employment

From 2003 to 2014, the proportion of temporary employment in

construction is higher than that in industries otherwise. In 2014, 15.5% of

employees in the construction industry were temporary workers, with 46%

6

increase compared with 10.6% in 2003, while only 9.1% of the workforce

is temporary employment in non-construction industry, with 23% increase

compared with 7.4% in 2003. The demographics of the temporary workers

were also distinct from the regular employees. In the 2011-2014 average,

temporary workers tended to constitute higher percentage at the age group

of 16-34 (temporary 35.8% versus regular 29.5%), education lower than

high school (temporary 46.6% versus regular 19.3%), Hispanic (temporary

44.3% versus regular 21.0%) and foreign-born (temporary 40.6% versus

regular 19.2%) (CPWR, 2018).

2.1.2 Health Status of Construction Workers

Body Mass Index (BMI)

Caban et al. (2005) reported that the prevalence rate of obesity among male

construction and extractive trades and construction labors were 18.45%

and 22.32% based on the National Health Interview Surveys from 1997 to

2002. CPWR (2016) reported that, between 2012 and 2014, 74.5%

construction workers were either overweight or obese, while 65.2% of the

overall workforce were at the same condition. Within construction, rates of

an abnormal BMI also increased with age. Workers with unhealthy BIM

constituted more than 80% of those at age group 55 years or older and 66%

of those at age group 35 years and younger (CPWR, 2016).

Diseases and self-related health condition

In a ten-year follow up of 14474 male construction workers, Arndt et al.,

(2005) found that construction workers had significantly higher risk of

disability resulting from cancer, respiratory and musculoskeletal diseases.

CPWR (2018) reported that 50.1% of construction workers had at least one

doctor-diagnosed health conditions, which escalated with age, ranging

from 26.4% at age group 16-34 to 87.5% at the age group of 65 and older.

In a study of construction bricklayers and supervisors, Boschman et al.,

(2013) found that 18% bricklayers and 20% supervisors rated themselves

as depressed, with which high work speed and quantity were associated

with. CPWR (2016) reported about 44% of construction workers rated their

7

mental health to be excellent, compared to 28% in terms of self-related

physical health. On the other hand, only 3.3% of construction workers

rated their mental health to be fair/poor while 9.8% in terms of physical

health. In addition, both self-reported physical and mental health worsened

after the age of 45. However, although construction workers tend to rate

their mental health more positively, it is plausible that the overestimation is

caused by lower level of awareness in terms of mental health issues

compared with physical health problems.

2.2 Mental health

2.2.1 Potential covariates

Mental health is found to be associated with general well-being through different

studies (CDC, 2016), whose potential covariates are also identified in different

studies.

Roberts et al. (2011) adjusted conditioning variables and found that commuting

time had a significant negative effect on psychological health by analyzing data

from the British Household Panel Survey, 1991-2004. In their study, apart from

including commuting time as the key explanatory variable of interest, they also

included net household income, housing quality and job satisfaction.

In the study of prevalence of current depression (CD) and frequent mental

distress (FMD) in Washington State, Fan et al. (2012) adjusted socio-

demographic as well as lifestyle behaviors factors. Depression is defined as a

common but serious mood disorder and is a subcategory of distress. People with

mental distress can experience problems in the way they think, feel or behave.

(National Institute of Mental Health, 2016).

2.2.2 Potential pathway towards negative effect on health

In the study of 383 Spanish workers, Silla et al. (2005) found temporary workers

displayed significant lower life satisfaction and well-being compared with

permanent workers. In the review of the association between temporary

employment and health, Virtanen et al. (2005) identified six potential pathway

towards negative effects on health and they were 1) erosion of income, 2) job

8

insecurity, 3) deficient benefits, 4) on-the-job-training, 5) lack of prospects for

promotion, and 6) exposure to hazardous work conditions. Based on the 27

studies they reviewed, they concluded that there was an association between

temporary employment and psychological morbidity.

2.3 Mental health in temporary construction workers

Fan et al. (2012) reported the prevalence of FMD was 8.1% in construction workers

compared to 7.5% in all workers in Washington State. However, the study did not

differentiate the temporary construction workers from permanent construction

workers. It is plausible that temporary employment is associated with lower mental

health performance in the construction industry, since temporary workers tend to have

lower income, less job security, insufficient benefits as well as higher exposure to

hazardous work conditions.

Kessler Index (K6) was verified to be valid to assess a person's non-specific

psychological distress during the past 30 days and was firstly included in the US

National Health Interview Survey in the 1997 (Kessler et al. 2002). The interviewees

are asked to rate the level of 1) felt nervous, 2) felt hopeless, 3) felt restless or fidgety,

4) felt so sad that nothing could cheer the person up, 5) felt everything was an effort

and 6) felt worthless during the past 30 days. Every question has a scale from none of

the time (0 score) to all of the time (4 scores) and the total scores range from 0-24.

An OR is a measure of association between an exposure and an outcome and can be

used in case-control as well as cross-section studies (Szumilas, 2010). Having a K6

score ≥13 is defined as having serious mental illness (SMI) (Kessler et al.1996). The

OR of SMI between permanent and temporary construction workers, for example,

represents the odds of having SMI when being a permanent construction worker,

compared to the odds of having SMI when not being a permanent construction

worker, namely being a temporary construction worker.

This study will look into the effect of temporary employment on the K6 score and the

odds of SMI with covariates adjusted.

9

3 Methodology

In order to determine the impact of temporary employment, K6 scores among temporary

construction workers will be compared with those among non- and permanent construction

workers. Potential covariates of demographics, income, housing, health status, lifestyle

behaviors and health insurance coverage will be adjusted in the analysis.

Both Medical Expenditure Panel Survey (MEPS) and Behavioral Risk Factor Surveillance

System-Washington State (BRFSS-WA) collect information on K6 score. As for the status

of temporary employment, MEPS is the only resource which has information on it. Due to

the gaps of information collected in these two major sources, the comparison of K6 scores

will be different at the national level as well as the state level. At the national level, the

comparisons of K6 scores are conducted among non-construction workers, permanent

construction workers and temporary construction workers. At the state level, the comparison

is conducted between non-construction workers and construction workers. The adjusted

covariates are the same except that housing status and binge drinking are not included in the

national level comparisons.

3.1 Available data sources

3.1.1 MEPS

MEPS, initiated in 1996, is a large-scale set of surveys and contains two major

components (AHRQ, 2009). One is the household component which gathers

information of the civilian non-institutional population of the United States, with

annual sample size around 15,000 households. The information of temporary

employment is collected in the household component. The other is the insurance

component which covers information on the health insurance plans offered by the

public and private employers. At the beginning of a year, a new panel of

households is selected and all the households will participate in five rounds of

interview in the current and subsequence year. As for one full calendar year,

information is available from six rounds of interview from two successive panels:

the third, the fourth and the fifth rounds from the panel initiated in previous year

and the first, the second and the third from the panel initiated in current year. In

MEPS, some variables, for instance industry and occupation, are asked in very

round of the interview and may change over the period of one full calendar year.

Since the information of K6 score is only collected in the fourth round of the

10

previous panel and the second round of the current panel, information of round

specific variables is derived from the same rounds from the previous and current

panel in this study. For variables which are collected only in specific round, the

information of them is used to represent the whole year.

3.1.2 BRFSS-WA

BRFSS-WA collects information of health changes of people in the Washington

State and is conducted by the Washington State Department of Health (WSDOH),

partnering with the Center for Disease Control and Prevention (CDC). More than

1000 interviews are conducted with people aged 18 years or older through

telephone every month (WSDOH, 2016). Apart from the core questionnaire

required by CDC, BRFSS-WA also implements state-specific modules to collect

other information. The information of temporary employment is not covered by

BRFSS-WA.

3.1.3 Available information and potential gaps

Demographics

The demographic data of interest are age, gender, race and ethnicity group,

marital status and education attainment, which are collected by both MEPS

and BRFSS-WA for the overall industries and construction industry. The

same information for the temporary construction workers is only covered

by MEPS.

Income

The income data of interest are annual personal wage, annual family

income and poverty status. MEPS collects information on these three

topics which covers the non-, permanent and temporary construction

workers. BRFSS-WA only collects information on annual family income,

which covers non-construction and construction workers.

Housing status

The housing status of interest are whether one owns or rents his/her home

and the types of building he/she lives in. BRFSS-WA is the only source for

this information which is not covered until 2008.

11

Occupation

The occupational data are available in both MEPS and BRFSS-WA for the

overall industries and construction industry, but MEPS is the only resource

which can distinguish temporary construction workers from the overall

construction workers. In addition, occupation variable is round specific in

MEPS, thus the information used in this study is derived from the fourth

round from previous year’s panel and the second round from current year’s

panel.

Health status and other covariates

The health status data of interest are the perceived general health, obesity,

had coronary heart disease (CHD), had stroke and had asthma. Other

covariates of interest are current smoking status, binge drinking, physically

inactivity, and health insurance coverage. MEPS collects information on all

the factors above except binge drinking for non-, permanent and temporary

construction workers. Perceived general health, obesity and physically

inactivity variables are round-specific in MEPS, thus the information of

them is based on the fourth round from previous year’s panel and the

second round from current year’s panel. BRFSS-WA has available

information for all factors for non-construction and construction workers.

K6 score

K6 score was not collected by MEPS until 2004 and not by BRFSS-WA

until 2007. The K6 score information of temporary construction workers is

only available in MEPS.

Temporary employment

The information on temporary employment status is only collected in

MEPS and is round-specific. Thus employment status is based on the

fourth round from previous year’s panel and the second round from current

year’s panel.

12

The availability of the information mentioned above is illustrated in Table 1 and

Table 2 at national level and Washington State level.

13

Table 1 Date Availability at National Level

Factors

Populations

Demographics Income Housing Health Status Other Covariates

(Except Binge

Drinking)

K6 Score

Annual Personal Wage/ Family

Income/ Poverty Status

Perceived General Health/

Obesity/ CHD/ Stroke/ Asthma

Non-construction

Workers MEPS MEPS − MEPS MEPS MEPS

Permanent

Construction

Workers

MEPS MEPS − MEPS MEPS MEPS

Temporary

Construction

Workers

MEPS MEPS − MEPS MEPS MEPS

14

Table 2 Date Availability at WA Level

Factors

Populations

Demographics Income Housing Status Health Status Other Covariates

(Include Binge

Drinking)

K6 Score

Annual Family

Income

Perceived General Health/

Obesity/ CHD/ Stroke/ Asthma

Non-construction

Workers BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA

Permanent

Construction

Workers

BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA BRFSS-WA

Temporary

Construction

Workers

− − − − −

15

3.2 Proposed models

At national level, the mean K6 score and the odds of SMI will be compared among the

non-, permanent and temporary construction labor force with demographics, income,

perceived general health, obesity, had CHD, had stroke, had asthma, current smoking

status, physically inactivity, and health insurance coverage adjusted.

At Washington State level, the mean K6 score and the odds of SMI will be compared

between the non-construction and construction labor force, with demographics,

income, housing status, perceived general health, obesity, had CHD, had stroke, had

asthma, current smoking status, binge drinking, physically inactivity, and health

insurance coverage adjusted.

In order to compare the mean K6 score between two different groups, linear

regression is employed. By using linear regression with all the covariates adjusted in

the model, the potential confoundings can be teased out in order to identify the

association between the temporary employment and the K6 score. In the linear

regression, t tests are conducted to test whether the mean K6 scores are significant

different between the compared two groups, permanent construction workers vs.

temporary construction workers for instance. F-tests are conducted to test whether the

exploratory variable and the covariates are significantly associated with the K6 score.

Logistic regression is employed to compare the odds of SMI since the dependent

variable (having SMI vs. not having SMI) is dichotomous. Furthermore, in order to

control the confoundings, all the covariates are included in the logistic model as

independent variables. In the logistic regression, Wald tests are conducted to test

whether the odd of SMI are significant different between the compared two groups,

permanent construction workers vs. temporary construction workers for instance. F-

tests are conducted to test whether the exploratory variable and the covariates are

significantly associated with the SMI.

16

4 Analysis and Results

4.1 At National Level

4.1.1 Data Extraction

At national level, all variable information is extracted from MEPS’s yearly

consolidated data files from 2008 to 2015 and some variables are computed based

on the original information for the purpose of this analysis. A response is treated

as having missing value when the answer to the question is “Not Ascertained”,

“Don’t know”, “Refused”, or “Inapplicable”.

A construction worker is defined as the one who reports his/her occupation as

“Construction, Extraction and Maintenance”. The rest of the workers are

considered as non-construction workers and their occupations include: 1)

Management, Business and Financial Operations, 2) Professional and Related, 3)

Service, 4) Sales and related, 5) Office and Administrative Support, 6) Farming,

Fishing, Forestry, 7) Installation, Repair, and Maintenance, 8) Production, 9)

Transportation and Material Moving and 10) Military Specific Occupations. The

distribution of the occupations is presented in Figure 1 and 2. Within the group of

construction workers, a temporary construction worker is defined as the one who

reports his/her current main job as either “temporary” or “seasonal”. The rest of

the group are considered as permanent construction workers.

Figure 1 Distribution of all the occupations (At national level, 2008-2015)

14358

22454

24743

10867

14306

1240

10270

16087

4611605

0

5000

10000

15000

20000

25000

30000

1 2 3 4 5 6 7 8 9 10

17

Figure 2 Distribution of Non-construction vs. Construction (At national level, 2008-2015)

The K6 score is collected as integers ranging from 0 to 24 and for the purpose of

the analysis, it is categorized into two groups: the group with serious mental

illness (SMI) whose K6 score is greater or equal to 13 points and the group

without SMI whose K6 score is less than 13.

The mean K6 score and the prevalence of SMI among different occupations are

presented in Table 3 and Table 4.

Table 3 Mean K6 Score among Different Occupations (At national level, 2008-2015)

Occupations N Mean Std Dev

1). Management, business, and financial operations 14358 2.43 3.18

2). Professional and related occupations 22454 2.63 3.26

3). Service occupations 24743 2.98 3.95

4). Sales and related occupations 10867 2.98 3.88

5). Office and administrative support 14306 2.83 3.73

6). Farming, fishing and forestry 1240 2.59 3.72

7). Construction, extraction, and maintenance

Permanent 8791 2.35 3.45

Temporary 1411 2.47 3.76

8). Production, transportation, material moving. 16087 2.56 3.71

9). Military specific occupations 461 1.97 3.23

10). Unclassified occupations 1605 2.36 3.74

106121

10270

0

20000

40000

60000

80000

100000

120000

Non-construction Construction, Extraction and Maintenance

18

The mean K6 score among permanent construction workers is only higher than

that among military specific occupations while the mean K6 score among

temporary construction workers is higher than that among management, business,

and financial operations, military specific occupations and unclassified

occupations.

Table 4 Prevalence of SMI among Different Occupations (At national level, 2008-2015)

Occupations Non-SMI SMI Prevalence of SMI

1). Management, business, and financial operations 12380 222 1.76%

2). Professional and related occupations 19563 379 1.90%

3). Service occupations 20361 753 3.57%

4). Sales and related occupations 9070 333 3.54%

5). Office and administrative support 12170 374 2.98%

6). Farming, fishing and forestry 1025 27 2.57%

7). Construction, extraction, and maintenance

Permanent 7340 168 2.24%

Temporary 1169 38 3.15%

8). Production, transportation, material moving. 13386 389 2.82%

9). Military specific occupations 116 3 2.52%

10). Unclassified occupations 1100 29 2.57%

As for the prevalence of SMI, permanent construction workers have a higher rate

than 1) management, business, and financial operations, and 2) professional and

related occupations. Temporary construction workers have a higher rate than all

the other occupations except 3) service occupations, and 4) sales and related

occupations.

The range of covariate age is from 18 years old to 65 years old and it is

categorized into three levels-18 to 24, 25 to 44 and 45 to 65 in this analysis. The

covariate race in 2013, 2014 and 2015 is rearranged into six categories in order to

be consistent with the information collected from 2008 to 2012, by combing

“Asian Indian”, “Chinese” and “Filipino” into one category “Asian”. The

marriage status is divided into two categories “Married” and “Not married”

which combines the “Widowed”, “Divorced”, “Separated” and “Never married”.

The highest education degree is collected from 2008 to 2015, except in 2013 and

2014. So the information for these two years is extracted from a similar variable

collected. However, there is still inconsistency in these two years, especially at

19

the levels of Master’s, Doctorate and professional degree, which are collapsed

into one level at 2013 and 2014, but separated in other years.

The personal annual wages and family annual income are categorized into five

groups-less than 25k, 25k-50k, 50k-75k, 75k-100k and more than 100k.

The information of BMI index is divided into four groups: Underweight (BMI

<18.5), Normal weight (18.5≤BMI≤24.9), Overweight (25≤BMI≤29.9) and

Obesity (BMI≥30).

The information for gender, ethnicity, poverty status, perceived general health,

physical activity, had stroke, had asthma and health insurance coverage is

extracted from the source without any other computation. As for physical

activity, there is minor difference in terms of the definition of physically active in

before and after 2011. It is defined as spending half hour or more in moderate to

vigorous physical activity at least three times a week from 2008 to 2010 and is

revised into at least five times in a week in 2011 and applied since then.

4.1.2 Descriptive Analysis

At national level, the distributions of all the covariates among non-, permanent

and temporary construction workers stay consistent from 2008 to 2015, except

annual personal wage and health insurance coverage. Thus, the overall summary

of all the variables is presented in Table 5 and the yearly summaries of 2008,

2012 and 2015 are included in the appendix Table 11, Table 12 and Table 13.

As for the distribution of age (Figure 3), people aged from 28 years old to 44

years old contribute to higher percentage of the temporary construction labor

force, which is 71%, compared with 60% of permanent construction and 60% of

non-construction labor force.

20

Figure 3 Distribution of age (At national level, 2008-2015)

As for the distribution of gender, it is skewed in construction industry regardless

of the status of employment compared with other industries as a whole. Female

workers only contribute to 3% of the overall construction labor force.

As for the distribution of race, there is little difference between the permanent

and the temporary construction labor force. In non-construction workers, white

people contribute a lower proportion while black people and Asians contribute a

larger proportion, with 69%, 19% and 7% respectively, compared with 83%, 11%

and 1% in permanent construction workers. As for the ethnic group, construction

industry has a higher percentage of Hispanic workers compared with the rest

industries as a whole. Hispanic workers constitute to 62% of the temporary

construction labor force, compared with 40% and 27% of the permanent and non-

construction labor force.

As for the marriage status, permanent construction workers have the highest rate

of being married (61%), while the non-construction workers go as the second

(52%) and the temporary construction workers as the last (45%).

13%

47%

40%

Non-construction

8%

52%

39%

Permanent Construction

14%

57%

29%

Temporary Construction

18-24

25-44

45-65

21

As for the highest education degree obtained (Figure 4), temporary construction

workers have the highest percentage of no degree obtained (46%) compared with

the permanent construction workers (27%) and non-construction workers (15%).

83% of the permanent and 91% of the temporary construction workers have a

highest degree of high school or below.

Figure 4 Distribution of education (At national level, 2008-2015)

As for the annual personal wage (Figure 5), temporary construction workers tend

to have the highest percentage in the annual personal wage category less than

25k. On average, 63% temporary construction workers fall into this category,

compared with 38% in permanent construction workers and 44% non-

construction workers. Meanwhile, the percentage of workers in this category

decreases over the years, from 63% in 2008 to 54% in 2015 within temporary

construction workers, from 41% in 2008 to 31% in 2015 within permanent

construction workers, and from 45% in 2008 in to 42% in 2015 within non-

construction workers (Figure 6).

15%

46%

38%

Non-construction

27%

57%

16%

Permanent Construction

46%

45%

9%

Temporary Construction

No Degree

High School or GED

Higher Degree

22

45% 44% 42%

41%36%

31%

63%

72%

54%

0%

10%

20%

30%

40%

50%

60%

70%

80%

2008 2012 2015

<25k (Non-construction) <25k (Permanent construction)

<25k (Temporary construction)

Figure 5 Distribution of annual personal wage (At national level, 2008-2015)

Figure 6 Non- vs. permanent vs. temporary construction workers with <25k annual personal wage

(At national level, 2008-2015)

Temporary construction workers also tend to have lower annual family income

that 69% of them have a family income less than 50K, compared with 49% in

permanent construction workers and 44% in non-construction workers (Figure 7).

44%

32%

13%

11%

Non-construction

38%

41%

14%

7%

Permanent Construction

63%

28%

6% 2%

Temporary Construction

<25k

25k-50k

50k-75k

>75k

23

The trend of having lower income in temporary construction workers also hold

when it comes to the poverty status, which is a relative measurement of the

annual family income. 52% of the temporary construction workers are considered

as low income or below, compared with 33% and 31% in the permanent

construction workers and non-construction workers respectively.

Figure 7 Distribution of annual family income (At national level, 2008-2015)

As for the perceived general health (Figure 8), the ratings of fair or poor consist

of a higher proportion in the temporary construction workers (11%) compared

with the permanent construction workers (9%) and the non-construction workers

(8%). Based on the BMI index, 77% of the temporary and 76% of the permanent

construction labor force are overweight or obese, while it is 66% for the rest of

the industries. In terms of smoking (Figure 9), 23% temporary construction

workers are considered as current smokers, compared with 26% and 16%

permanent and non-construction workers. Permanent construction workers are

considered to be more physically active (61%) and the rates are very close in the

temporary construction workers (53%) and the non-construction workers (55%).

As for other health status, distributions are similar within the three groups, except

permanent (5%) and temporary (4%) construction workers have lower asthma

rate compared with the non-construction workers (8%).

18%

26%

20%

36%

Non-construction

18%

31%22%

29%

Permanent Construction

37%

32%

16%

15%

Temporary Construction

<25k

25k-50k

50k-75k

>75k

24

Figure 8 Distribution of perceived general health (At national level, 2008-2015)

Figure 9 Distribution of current smoking status (At national level, 2008-2015)

27%

36%

29%

7%1%

Non-construction

25%

34%

31%

8%

1%

Permanent Construction

31%

28%

31%

10%1%

Temporary Construction

Excellent

Very good

Good

Fair

Poor

16%

84%

Non-construction

26%

74%

Permanent Construction

23%

77%

Temporary Construction

Current Smoker

Current Non-Smoker

25

80%

20%

Non-construction

64%

36%

Permanent Construction

39%

61%

Temporary Construction

Covered

Uninsured

As for the health insurance coverage (Figure 10), the uninsured rate is higher in

the construction industry, with 36% permanent and 61% temporary construction

workers uninsured, while the uninsured rate is only 20% in the non-construction

industries. Meanwhile, a significant decrease in terms of uninsured rate is

identified among temporary construction workers, which decreases from 70% in

2008 to 50% in 2015. The health insurance coverage rate from 2008 to 2015 is

presented in Figure 11.

Figure 10 Distribution of health insurance coverage (At national level, 2008-2015)

Figure 11 Health insurance coverage in non- vs. permanent vs. temporary construction workers (At

national level, 2008-2015)

70%65%

50%

0%

20%

40%

60%

80%

100%

2008 2012 2015

Covered (Non-construction) Covered (Permanent construction)

Covered (Temporary construction) Uninsured (Non-construction)

Uninsured (Permanent construction) Uninsured (Temporary construction)

26

Table 5 Descriptive Summary of All the Variables (At national level, 2008-2015)

Non-

construction %

Permanent

construction %

Temporary

construction %

Age

28-24 13433 13% 717 8% 195 14%

25-44 50363 47% 4614 52% 804 57%

45-65 42325 40% 3460 39% 412 29%

Gender

Male 49875 47% 8549 97% 1372 97%

Female 56246 53% 242 3% 39 3%

Race

White 68606 69% 6889 83% 1101 85%

Black 19118 19% 910 11% 128 10%

American Indian/Alaska

Native 839 1% 72 1% 15 1%

Asian 7236 7% 203 2% 24 2%

Native Hawaiian/Pacific

Islander 1440 1% 55 1% 8 1%

Multiple races reported 2046 2% 123 1% 20 2%

Ethic group

Hispanic 28858 27% 3509 40% 880 62%

Not Hispanic 77263 73% 5282 60% 531 38%

Marriage status

Married 55514 52% 5354 61% 633 45%

Not married 50603 48% 3437 39% 778 55%

Highest degree

No Degree 14161 15% 2078 27% 543 46%

GED 3276 4% 470 6% 59 5%

High School Diploma 39025 43% 3841 50% 474 40%

Bachelor's Degree 17497 19% 402 5% 38 3%

Master's Degree 7684 8% 49 1% 8 1%

Doctorate Degree 1393 2% 7 0% 1 0%

Other Degree 8592 9% 766 10% 61 5%

Annual personal wage

<25k 46680 44% 3299 38% 895 63%

25k-50k 33798 32% 3592 41% 398 28%

50k-75k 14164 13% 1263 14% 89 6%

75k-100k 5950 6% 437 5% 21 1%

>100k 5529 5% 200 2% 8 1%

Annual family income

<25k 19106 18% 1567 18% 526 37%

27

25k-50k 27588 26% 2764 31% 447 32%

50k-75k 20881 20% 1929 22% 224 16%

75k-100k 14236 13% 1250 14% 104 7%

>100k 24310 23% 1281 15% 110 8%

Poverty status

Poor/Negative 10311 10% 879 10% 301 21%

Near poor 4893 5% 451 5% 123 9%

Low income 16767 16% 1569 18% 313 22%

Middle income 35801 34% 3364 38% 483 34%

High income 38349 36% 2528 29% 191 14%

Perceived general health

Excellent 28344 27% 2228 25% 439 31%

Very good 37901 36% 2992 34% 390 28%

Good 30920 29% 2746 31% 434 31%

Fair 7795 7% 729 8% 137 10%

Poor 1055 1% 93 1% 10 1%

Obesity

Under weight 1444 1% 55 1% 9 1%

Normal 33867 33% 2018 24% 310 23%

Overweight 36184 35% 3734 44% 672 50%

Obesity 31767 31% 2719 32% 358 27%

Smoking status

Current smoker 14894 16% 1985 26% 283 23%

Current non-smoker 77922 84% 5630 74% 936 77%

Physical activity

Physically active 57173 55% 5300 61% 742 53%

Physically inactive 47502 45% 3402 39% 655 47%

CHD

Had CHD 1766 2% 199 2% 29 2%

Not had CHD 104270 98% 8589 98% 1381 98%

Stroke

Had stroke 1170 1% 70 1% 7 0%

Not had stroke 104880 99% 8719 99% 1403 100%

Asthma

Had asthma 8520 8% 428 5% 63 4%

Not had asthma 97538 92% 8360 95% 1347 96%

Health insurance

Covered by any private

insurance 75981 72% 5180 59% 404 29%

Covered by public

insurance only 9136 9% 457 5% 142 10%

Uninsured 21004 20% 3154 36% 865 61%

28

4.1.3 Inferential Analysis

The linear regression model is used to compare the mean of K6 score among the

non-, permanent and temporary construction labor force. The logistic regression

model is used to compare the odd of SMI among the three groups mentioned

above. In both models, all the covariates are adjusted as factors and the results are

presented in Table 6. Since all explanatory variables are treated as categorical

variables in the analysis, the first category of these variables is considered as the

control group with which the rest categories are compared. Thus results of the

linear and logistic regression are not presented for these controls group in the

result section.

29

Table 6 Mean K6 Score and OR of SMI among three groups with covariates adjusted (At national level, 2008-2015).

K6 Score OR of Serious Mental Illness

Est 95%CI Pr(>|t|) Pr(>F) OR 95%CI Pr(>|z|) Pr(>F)

Occupation <0.001

<0.001

Non-construction

Permanent Construction -0.24 (-0.335, -0.152) <0.001 0.81 (0.662, 0.983) 0.033

Temporary Construction -0.06 (-0.304, 0.177) 0.604 1.11 (0.718, 1.717) 0.639

Year <0.001 <0.001

2008

2009 -0.02 (-0.107, 0.069) 0.667 1.00 (0.859, 1.165) 0.992

2010 -0.18 (-0.270, -0.088) <0.001 0.98 (0.837, 1.151) 0.821

2011 -0.21 (-0.296, -0.118) <0.001 0.89 (0.764, 1.047) 0.166

2012 -0.28 (-0.384, -0.176) <0.001 0.99 (0.825, 1.191) 0.928

2013 -0.37 (-0.479, -0.265) <0.001 0.82 (0.676, 1.005) 0.056

2014 -0.62 (-0.706, -0.532) <0.001 0.64 (0.539, 0.764) <0.001

2015 -0.53 (-0.631, -0.422) <0.001 0.66 (0.536, 0.818) <0.001

Age <0.001 <0.001

18-24

25-44 -0.21 (-0.300, -0.123) <0.001 0.90 (0.777, 1.052) 0.193

45-65 -0.53 (-0.621, -0.438) <0.001 0.71 (0.604, 0.833) <0.001

Gender <0.001 <0.001

Male

Female 0.44 (0.393, 0.496) <0.001 1.41 (1.278, 1.564) <0.001

Race <0.001 0.002

White

Black -0.64 (-0.710, -0.569) <0.001 0.78 (0.685, 0.883) <0.001

American Indian/Alaska Native -0.26 (-0.516, -0.009) 0.042 0.79 (0.498, 1.246) 0.308

30

Asian -0.24 (-0.340, -0.147) <0.001 0.84 (0.665, 1.057) 0.136

Native Hawaiian/Pacific Islander 0.00 (-0.237, 0.227) 0.968 1.46 (0.997, 2.139) 0.052

Multiple races reported 0.00 (-0.191, 0.183) 0.963 1.00 (0.736, 1.361) 0.996

Ethic group <0.001 0.005

Hispanic

Not Hispanic 0.48 (0.410, 0.545) <0.001 1.21 (1.072, 1.375) 0.002

Marriage status <0.001 <0.001

Married

Not married 0.25 (0.191, 0.305) <0.001 1.16 (1.045, 1.287) 0.005

Highest degree (0.000, 0.000) <0.001 (0.000, 0.000) <0.001

No Degree (0.000, 0.000) (0.000, 0.000)

GED 0.30 (0.135, 0.462) <0.001 1.36 (1.108, 1.658) 0.003

High School Diploma 0.08 (0.000, 0.165) 0.051 1.04 (0.911, 1.182) 0.578

Bachelor's Degree 0.25 (0.155, 0.350) <0.001 0.87 (0.723, 1.050) 0.148

Master's Degree 0.30 (0.184, 0.408) <0.001 0.73 (0.553, 0.962) 0.026

Doctorate Degree 0.34 (0.161, 0.512) <0.001 1.07 (0.664, 1.732) 0.774

Other Degree 0.19 (0.076, 0.296) <0.001 1.11 (0.915, 1.340) 0.296

Annual family income <0.001 <0.001

<25k

25k-50k -0.32 (-0.405, -0.234) <0.001 0.83 (0.733, 0.936) 0.002

50k-75k -0.48 (-0.568, -0.385) <0.001 0.76 (0.658, 0.882) <0.001

75k-100k -0.59 (-0.685, -0.486) <0.001 0.60 (0.499, 0.731) <0.001

>100k -0.76 (-0.851, -0.662) <0.001 0.49 (0.401, 0.589) <0.001

Perceived general health <0.001 <0.001

Excellent

Very good 0.65 (0.600, 0.704) <0.001 1.61 (1.349, 1.918) <0.001

Good 1.47 (1.401, 1.529) 3.22 (2.724, 3.810) <0.001

31

Fair 3.34 (3.210, 3.479) 9.17 (7.658, 10.990) <0.001

Poor 6.21 (5.756, 6.671) <0.001 26.65 (21.007, 33.800) <0.001

Obesity 0.055 <0.001

Under weight

Normal -0.04 (-0.280, 0.191) 0.712 0.90 (0.622, 1.312) 0.592

Overweight -0.08 (-0.322, 0.152) 0.483 0.87 (0.601, 1.271) 0.480

Obesity -0.13 (-0.373, 0.105) 0.271 0.95 (0.654, 1.378) 0.785

Smoke status <0.001 <0.001

Current smoker

Current non-smoker -0.70 (-0.777, -0.623) <0.001 0.58 (0.518, 0.640) <0.001

Physical activity <0.001 <0.001

Physically active

Physically inactive 0.22 (0.169, 0.269) <0.001 1.19 (1.083, 1.303) <0.001

CHD 0.010 0.051

Had CHD

Not had CHD -0.29 (-0.511, -0.071) 0.009 0.78 (0.609, 1.001) 0.051

Stroke <0.001 <0.001

Had stroke

Not had stroke -0.54 (-0.842, -0.230) <0.001 0.62 (0.476, 0.820) <0.001

Asthma <0.001 <0.001

Had asthma

Not had asthma -0.48 (-0.578, -0.372) <0.001 0.74 (0.649, 0.852) <0.001

Health insurance 0.000 0.520

Covered by any private insurance

Covered by public insurance only 0.23 (0.121, 0.349) <0.001 1.07 (0.916, 1.250) 0.394

Uninsured -0.01 (-0.085, 0.062) 0.753 0.98 (0.860, 1.106) 0.697

32

As for the K6 score, permanent construction workers have a significantly lower

(p<0.001) score than the non-construction workers with other covariates adjusted.

The mean K6 score among permanent construction workers is 0.24 point lower

than that among the non-construction workers (95% CI: -0.335,-0.152).

Temporary construction workers are estimated to have a 0.06-point lower K6

score compared with the non-construction workers with all covariates adjusted,

however, the difference is not significant (95%CI: -0.304, 0.177). A comparison

between permanent and temporary construction labor force is also performed.

The mean K6 score among temporary construction workers is 0.18-point higher

than that among permanent construction workers, but the difference is not

significant (95% CI: -0.072, 0.431).

As for the covariate age, those among 25-44 or 45-65 tend to have a significantly

lower K6 score (p<0.001). People from 25-44 have a mean K6 score 0.21-point

lower than those from 18-24 (95%CI: -0.300,-0.123) and people from 45-65 have

a mean K6 score 0.53-point lower than those from 18-24 (95%CI: -0.621, -

0.438).

As for gender, female tend to have a significantly higher K6 score (p<0.001),

which is 0.44-point higher than that among male (95%CI: 0.393, 0.496). As for

the ethic group, the non-Hispanic tend to have a significantly higher K6 score

(p=0.002), which is 0.48-point higher that that among the Hispanic (95%CI:

0.410, 0.545).

As for the annual family income, the mean K6 score in the group of 25k-50k or

higher is significantly lower than that in the group with annual family income less

than 25k (p<0.05). The difference in the mean K6 score in these groups is 0.32-

point (95%CI: -0.405, -0.234), 0.48-point (95%CI: -0.568, -0.385), 0.59-point

(95%CI: -0.685,-0.486) and 0.76-point (95%CI: -0.851. -0.662) lower than the

lowest category.

As for smoking status, current non-smokers tend to have a significantly lower K6

score compared with the current smokers (p<0.001). It is estimated that the mean

33

K6 score of the former group is 0.70-point lower than that among the latter group

(95% CI: -0.777,-0.623).

For other covariates, they are also significantly associated with the mean K6

score, except obesity.

As for the prevalence of SMI, with all covariates adjusted, permanent

construction workers have a significantly lower odds of having SMI compared

with non-construction workers (p=0.033). The odds of SMI among permanent

construction workers is 81% of the odds among non-construction workers

(95%CI: 0.662, 0.983). The odds of SMI among temporary construction workers

is 1.11 times the odds among non-construction workers, but the difference is not

significant (95%CI:0.718,1.717). A companion between permanent and

temporary construction labor force is also performed with other covariates

adjusted. The odds among temporary construction workers is 1.38 times that

among permanent construction workers, however, the difference is not significant

(95% CI: 0.865, 2.192). In addition, all the covariates are significant associated

with the prevalence of SMI, except had CHD and health insurance coverage

status.

As for the covariate age, only those among 45-65 tend to have a significantly

lower odds of SMI (p<0.001). The odds among people from 45-65 is 71% of the

odds among those from 18-24 (95%CI: 0.604, 0.833).

As for gender, female tend to have a significantly higher odds of having SMI

(p<0.001), which is 1.41 times the odds among male (95%CI: 1.278, 1.564). As

for the ethic group, the non-Hispanic have a significantly higher odds of having

SMI (p=0.002), which is 1.21-times the odds among the Hispanic.

As for the annual family income, the odds of SMI in the group of 25k-50k or

higher is significantly lower than that in the group of less than 25k. The odds in

these groups is 83% (95%CI: 0.733, 0.936), 76% (95%CI: 0.658, 0.882), 60%

(95%CI: 0.499, 0.731) and 49% (95%CI: 0.401, 0.589) of the odds in the lowest

category.

34

As for smoking status, current non-smokers tend to have a significantly lower

odds compared with the current smokers (p<0.001). It is estimated that odds of

SMI in the former group is 58% of the odds in the latter group (95% CI: 0.518,

0.640).

For other covariates, they are also significantly associated with the SMI, except

the covariates had CHD and health insurance coverage.

4.2 At Washington State Level

4.2.1 Data Extraction

At Washington State level, all variable information is extracted from BRFSS-

WA’s data file from 2011 to 2016 and some variables are computed based on the

original information for the purpose of this analysis. A response is treated as

having missing value when the answer to the question is “Missing”, “Don’t

know”, “Refused”, or “Not sure”.

A construction worker is defined as the one who reports his/her occupation as

“Construction, and Extraction”. The rest of the workers are considered as non-

construction workers and their occupations include: 1) Management, Business

and Financial Operations, 2) Professional and Related, 3) Service, 4) Sales and

related, 5) Office and Administrative Support, 6) Farming, Fishing, Forestry, 7)

Construction and Extraction, 8) Installation, Repair, and Maintenance, 9)

Production, 10) Transportation and Material Moving and 11) Military Specific

Occupations. The distribution of the occupations is presented in Figure 12 and 13.

35

Figure 12 Distribution of all the occupations (At WA level, 2011-2016)

Figure 13 Distribution of Non-construction vs. Construction (At WA level, 2011-2016)

The K6 score is collected as integers ranging from 0 to 24. For the purpose of the

analysis, it is transformed in the same way as the MEPS data, which categorizes it

into SMI and non-SMI groups. The distribution of K6 score and SMI among

different occupations are presented in Table 7 and Table 8.

6145

10670

4358

2492

3414

584

1429937

1322 1576

161

0

2000

4000

6000

8000

10000

12000

1 2 3 4 5 6 7 8 9 10 11

31659

1429

0

5000

10000

15000

20000

25000

30000

35000

Non-construction Construction and Extraction

36

Table 7 Mean K6 Score among Different Occupations (At WA level, 2011-2016)

Occupation N Mean Std Dev

1). Management, Business & Financial Operations 6145 2.24 2.72

2). Professional and Related 10670 2.45 2.83

3). Service 4358 3.24 3.85

4). Sales and related 2492 2.76 3.38

5). Office and Administrative Support 3414 2.78 3.42

6). Farming, Fishing, Forestry 584 2.58 3.37

7). Construction and Extraction 1429 2.52 3.11

8). Installation, Repair, and Maintenance 937 2.27 2.96

9). Production 1322 2.80 3.52

10). Transportation and Material Moving 1576 2.50 3.33

11). Military Specific Occupations 161 2.61 3.69

The mean K6 score among permanent construction workers is higher than that

among 1) management, business and financial operations, 2) professional and

related, 8) installation, repair, and maintenance, and 10) transportation and

material moving.

Table 8 Prevalence of SMI among Different Occupations (At WA level, 2011-2016)

Non-SMI SMI Prevalence of SMI

1). Management, Business & Financial Operations 5392 55 1.01%

2). Professional and Related 9345 113 1.19%

3). Service 3551 141 3.82%

4). Sales and related 2123 51 2.35%

5). Office and Administrative Support 2927 73 2.43%

6). Farming, Fishing, Forestry 479 7 1.44%

7). Construction and Extraction 1198 21 1.72%

8). Installation, Repair, and Maintenance 779 7 0.89%

9). Production 1108 25 2.21%

10). Transportation and Material Moving 1308 27 2.02%

11). Military Specific Occupations 132 5 3.65%

As for the prevalence of SMI, construction workers have a higher rate than 1)

management, business, and financial operations, 2) professional and related

occupations, 6) farming, fishing, forestry, and 8) installation, repair, and

maintenance.

37

The range of covariate age is from 18 years old to 65 years old and it is

categorized into three levels-18 to 24, 25 to 44 and 45 to 65 in this analysis. The

marriage status is divided into two categories “Married” and “Not married”,

which combines the “Widowed”, “Divorced”, “Separated”, “Never married” and

“Member of unmarried couple”.

The smoking status is categorized into two levels-current smoker and current

non-smoker. The former group contains those who report current daily or

occasionally smoking and the latter contains those who report no current smoking

or never smoked.

The information for gender, race and ethnicity, highest education degree, annual

family income, perceived general health, physical activity, binge drinking, had

stroke, had asthma, health insurance coverage and housing status is extracted

from the source without any other computation.

4.2.2 Descriptive Analysis

At state level, the distributions of all the covariates among non-construction and

construction workers stay consistent from 2011 to 2016, except annual family

income, health insurance coverage and housing status. Thus, the overall summary

of the variables is presented in Table 9. The yearly summaries from 2011 to 2016

are included in the appendix from Table 14 to Table 19.

As for the distribution of age (Figure 14), people aged from 28 to 44 contribute to

a higher percentage of the construction labor force, which is 45% compared with

40% of non-construction labor force.

Figure 14 Distribution of age (At WA level, 2011-2016)

6%

34%

61%

Non-construction

7%

38%55%

Construction

18-24

25-44

45-65

38

As for the distribution of gender, it is skewed in construction industry compared

with other industries as a whole. Female workers only contribute to 7% of the

overall construction labor force.

As for the distribution of race, there is little difference between the construction

and non-construction labor force. As for the ethnic group, construction industry

has a higher percentage of Hispanic workers compared with the rest industries as

a whole, with 11% Hispanic workers in construction industry compared to 7% in

non-construction industries.

As for the marriage status, construction workers have a slightly lower marriage

rate of 57%, while the non-construction workers have a marriage rate of 61%.

As for the highest education obtained (Figure 15), majority of the construction

workers are high school graduates or have some college or technical school

education, which contributes to 76% of the construction labor force, compared

with 46% in the non-construction labor force.

Figure 15 Distribution of education (At WA level, 2011-2016)

As for the annual family income (Figure 16), construction workers have a higher

percentage in the annual family income category less than 50k. On average, 41%

construction workers fall into this category, compared with 33% in non-

construction workers. At the same time, a significant increase in the proportion of

the workers who has an annual family income greater than 75k can be identifies

0%

23%

77%

Non-construction

0%

51%

49%

Construction

No Degree

High School or GED

Higher Degree

39

in both groups. 31% construction workers and 42% non-construction workers fall

into this category in 2011 while 42% construction workers and 52% non-

construction workers in 2016. The annual family income from 2011 to 2016 is

presented in Figure 17.

Figure 16 Distribution of annual family income (At WA level, 2011-2016)

Figure 17 Annual family income non-construction vs. construction (At WA level, 2011-2016)

As for the perceived general health (Figure 18), the ratings of fair or poor consist

of a slightly higher proportion in construction workers (12%) compared with non-

construction workers (9%). Based on the BMI index, 70% of the construction

labor force is overweight or obese, while it is 64% for the rest of the industries. In

terms of smoking, construction workers have a higher rate with 23% construction

42%46% 47%

49% 51% 52%

31%28%

31%

39%42% 42%

0%

10%

20%

30%

40%

50%

60%

2011 2012 2013 2014 2015 2016

<50k (Non-construction) <50k (Construction)

50k-75k (Non-construction) 50k-75k (Construction)

>75k (Non-construction) >75k (Construction)

13%

21%

19%

48%

Non-construction

16%

25%

23%

36%

Construction

<25k

25k-50k

50k-75k

>75k

40

workers are current smokers, compared with 13% in non-construction workers

(Figure 19). Construction workers also have a higher proportion of heavy alcohol

consumption, with 11% compared to 7% in non-construction workers. In terms of

physical activity, construction workers are considered to be less physically active,

with a rate of 80% compared to 85% in the non-construction workers.

Figure 18 Distribution of perceived general health (At WA level, 2011-2016)

Figure 19 Distribution of current smoking status (At WA level, 2011-2016)

As for other health status, the distributions are similar in the two groups, except

construction workers seem to have lower asthma rate (10%) compared with the

non-construction workers (14%).

As for the health insurance coverage (Figure 20), the uninsured rate is higher in

the construction industry, with 23% construction workers uninsured, while the

uninsured rate is only 11% in the non-construction industries. Meanwhile, a

significant decrease in the rate of uninsured construction workers can be

22%

40%

29%

8%

1%

Non-construction

19%

33%37%

11%

1%

Construction

Excellent

Very good

Good

Fair

Poor

13%

87%

Non-construction

23%

77%

Construction

Current Smoker

Current Non-

Smoker

41

identified, which dropped from 27% in 2011 to 16% in 2016. The health

insurance coverage rate from 2011 to 2016 is presented in Figure 21.

Figure 20 Trend of health insurance coverage (At WA level, 2011-016)

Figure 21 Health insurance coverage rate in non-construction vs. construction (At WA level, 2011-

2016)

As for housing status (Figure 22), construction workers tend to have a higher rate

of renting their home rather than owning it. 31% construction workers rent their

home and 67% own their home, while 24% non-construction workers rent their

home and 74% own their home. Meanwhile a significant decrease in the

percentage of owning home is identified in both groups. 80% construction

workers and 78% non-construction workers own their home in 2011 while 62%

of the former group and 69% of the latter group own their home in 2016. The

housing status from 2011 to 2016 is presented in Figure 23.

89%

11%

Non-construction

77%

23%

Construction

Covered

Uninsured

88% 86% 85%92% 93% 92%

73%67% 68%

81% 83% 84%

12% 14% 15%8% 7% 8%

27%33% 32%

19% 17% 16%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

2011 2012 2013 2014 2015 2016

Covered (Non-construction) Covered (Construction)

Uninsured (Non-construction) Uninsured (Construction)

42

Figure 22 Distribution of housing status (At WA level, 2011-2016)

Figure 23 Trend of housing status in non-construction vs. construction (At WA level, 2011-2016)

As for type of building people live in, the information is not available in 2011.

Based on the data from 2012 to 2016, construction workers and non-construction

workers have very close distribution on average.

73%

24%

3%

Non-construction

67%

31%

3%

Construction

Own

Rent

Other arrangement

78%75%

72% 72% 70% 69%80%

65% 66%69%

63% 62%

19%22%

25% 25% 26%28%

19%

31% 32%29%

34% 35%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

2011 2012 2013 2014 2015 2016

Own (Non-construction) Own (Construction)

Rent (Non-construction) Rent (Construction)

Other (Non-construction) Other (Construction)

43

Table 9 Descriptive Summary of All the Variables (At WA level, 2011-2016)

Non-construction % Construction %

Age

18-24 1853 6% 96 7%

25-44 10640 34% 542 38%

45-65 19166 61% 791 55%

Gender

Male 14647 46% 1324 93%

Female 17012 54% 105 7%

Race

White 26861 87% 1199 86%

Black 628 2% 18 1%

American Indian/Alaska

Native 386 1% 27 2%

Asian 1169 4% 22 2%

Pacific Islander 185 1% 4 0%

Other race 1005 3% 85 6%

Multiple races reported 737 2% 41 3%

Ethic group

Hispanic 2359 7% 156 11%

Not Hispanic 29139 93% 1264 89%

Marriage status

Married 19236 61% 812 57%

Not married 12290 39% 609 43%

Highest degree

None or kindergarten 55 0% 1 0%

Grade 1-8 516 2% 43 3%

Grade 9-11 796 3% 89 6%

Grade 12 or GED 5811 18% 601 42%

College 1-3 years 8945 28% 482 34%

College 4 years or more 15505 49% 212 15%

Annual family income

<10k 445 2% 32 2%

10k-15k 594 2% 28 2%

15k-20k 1047 4% 69 5%

20k-25k 1642 6% 80 6%

25k-35k 2346 8% 114 9%

35k-50k 3696 13% 214 16%

50k-75k 5396 19% 300 23%

>75k 13812 48% 472 36%

Perceived general health

44

Excellent 6978 22% 266 19%

Very good 12597 40% 469 33%

Good 9279 29% 523 37%

Fair 2416 8% 155 11%

Poor 367 1% 15 1%

Obesity

Under weight 353 1% 6 0%

Normal 10331 35% 419 30%

Overweight 10780 36% 580 42%

Obesity 8237 28% 383 28%

Smoking status

Current smoker 3930 13% 326 23%

Current non-smoker 27165 87% 1067 77%

Heavy alcohol consumption

Yes 2274 7% 149 11%

No 28407 93% 1222 89%

Physical activity

Physically active 26394 85% 1113 80%

Physically inactive 4609 15% 281 20%

CHD

Had CHD 509 2% 30 2%

Not had CHD 31072 98% 1390 98%

Stroke

Had stroke 299 1% 19 1%

Not had stroke 31318 99% 1406 99%

Asthma

Had asthma 4367 14% 139 10%

Not had asthma 27177 86% 1282 90%

Health insurance

Covered 28193 89% 1087 77%

Uninsured 3367 11% 330 23%

Housing condition

Own 22917 73% 946 67%

Rent 7570 24% 436 31%

Other arrangement 976 3% 39 3%

Building Type (2012-2016)

Mobile home 1565 6% 125 10%

Detached one-family house 19963 77% 905 75%

Attached home (duplex/etc) 849 3% 53 4%

Building with 1 - 4 apts/condos 712 3% 37 3%

Building with 5 + apts/condos 2510 10% 76 6%

45

Apt/Condo, units unknown 33 0% 0 0%

Other 232 1% 8 1%

4.2.3 Inferential Analysis

The linear regression model is used to compare the mean of K6 score among the

non-construction and construction labor force. The logistic regression model is

used to compare the odd of having SMI among the two groups mentioned above.

In both models, all the covariates are adjusted as factors and the results are

presented in Table 10. For every categorical variable, the first category is

considered as the control group with which the rest categories are compared, thus

results related to the linear and logistic regression are not presented for these

control groups in the result section.

46

Table 10 Mean K6 Score and OR of SMI among two groups with covariates adjusted (At WA level, 2011-2016

K6 Score OR of SMI

Est 95% CI Pr(>|t|) Pr(>F) e(Est) 95% CI Pr(>|t|) Pr(>F)

Occupation 0.414 0.672

Non-construction

Construction -0.07 (-0.251, 0.103) 0.414 0.89 (0.532, 1.502) 0.672

Year <0.001 0.001

2011

2012 -0.02 (-0.141, 0.098) 0.721 1.69 (1.122, 2.553) 0.012

2013 0.08 (-0.054, 0.206) 0.254 1.83 (1.191, 2.807) 0.006

2014 0.22 (0.082, 0.358) 0.002 1.90 (1.213, 2.967) 0.005

2015 -0.03 (-0.147, 0.096) 0.676 1.25 (0.805, 1.957) 0.316

2016 -0.13 (-0.262, -0.003) 0.045 2.22 (1.454, 3.390) <0.001

Age <0.001 0.019

18-24

25-44 -0.45 (-0.681, -0.216) <0.001 0.72 (0.492, 1.067) 0.103

45-65 -0.98 (-1.211, -0.745) <0.001 0.58 (0.388, 0.867) 0.008

Gender <0.001 <0.001

Male

Female 0.33 (0.252, 0.404) <0.001 1.48 (1.183, 1.862) <0.001

Race 0.149 0.887

White

Black -0.15 (-0.420, 0.125) 0.288 0.75 (0.361, 1.550) 0.435

American Indian/Alaska Native 0.15 (-0.237, 0.541) 0.442 0.79 (0.364, 1.698) 0.540

Asian -0.25 (-0.452, -0.049) 0.015 0.81 (0.399, 1.654) 0.567

pacific Islander -0.29 (-0.834, 0.250) 0.291 0.35 (0.046, 2.592) 0.302

Other race -0.01 (-0.291, 0.277) 0.962 0.90 (0.461, 1.742) 0.746

47

Multiple races reported 0.11 (-0.160, 0.376) 0.430 0.95 (0.532, 1.707) 0.871

Ethic group 0.002 0.219

Hispanic

Not Hispanic 0.33 (0.126, 0.543) 0.002 1.43 (0.809, 2.514) 0.219

Marriage status <0.001 <0.001

Married

Not married 0.42 (0.339, 0.510) <0.001 1.65 (1.289, 2.125) <0.001

Highest degree <0.001 <0.001

None or Kindergarten

Grade 1-8 -1.15 (-2.811, 0.515) 0.176 112711.93 (30869.850, 411533.552) <0.001

Grade 9-11 0.12 (-1.529, 1.771) 0.886 310213.90 (118517.106, 811972.777) <0.001

Grad 12 or GED 0.15 (-1.480, 1.771) 0.861 272356.99 (108811.364, 681714.897) <0.001

College 1-3 years 0.35 (-1.271, 1.979) 0.669 325483.15 (128696.374, 823172.218) <0.001

College 4 years or more 0.50 (-1.130, 2.120) 0.551 278775.37 (109129.630, 712141.231) <0.001

Annual family income <0.001 <0.001

<10k

10k-15k -0.49 (-1.108, 0.137) 0.126 0.78 (0.421, 1.444) 0.429

15k-20k -0.68 (-1.258, -0.107) 0.020 0.68 (0.392, 1.179) 0.169

20k-25k -1.07 (-1.613, -0.530) <0.001 0.67 (0.392, 1.137) 0.137

25k-35k -1.34 (-1.865, -0.816) <0.001 0.48 (0.284, 0.806) 0.006

35k-50k -1.48 (-1.994, -0.961) <0.001 0.45 (0.269, 0.762) 0.003

50k-75k -1.70 (-2.208, -1.183) <0.001 0.30 (0.171, 0.522) <0.001

>75k -1.72 (-2.234, -1.211) <0.001 0.25 (0.141, 0.433) <0.001

Perceived general health <0.001 <0.001

Excellent

Very good 0.57 (0.496, 0.651) <0.001 2.33 (1.410, 3.861) <0.001

48

Good 1.23 (1.128, 1.329) <0.001 4.72 (2.867, 7.780) <0.001

Fair 2.86 (2.642, 3.072) <0.001 13.08 (7.757, 22.048) <0.001

Poor 4.97 (4.285, 5.660) <0.001 35.77 (19.639, 65.166) <0.001

Obesity <0.001 0.007

Under weight

Normal -0.52 (-1.001, -0.047) 0.031 0.36 (0.195, 0.666) 0.001

Overweight -0.65 (-1.130, -0.173) 0.008 0.33 (0.180, 0.622) <0.001

Obesity -0.70 (-1.186, -0.221) 0.004 0.35 (0.189, 0.656) <0.001

Smoke <0.001 <0.001

Smoke

Not smoke -0.65 (-0.790, -0.511) <0.001 0.60 (0.466, 0.763) <0.001

Heavy alcohol consumption <0.001 <0.001

Yes

No -0.63 (-0.784, -0.472) <0.001 0.55 (0.411, 0.748) <0.001

Physical activity 0.732 0.858

Physically active

Physically inactive -0.02 (-0.140, 0.098) 0.732 1.02 (0.798, 1.311) 0.858

CHD 0.216 0.167

Had CHD

Not had CHD 0.20 (-0.116, 0.515) 0.216 1.72 (0.798, 3.697) 0.167

Stroke 0.100 0.127

Had stroke

Not had stroke -0.42 (-0.923, 0.079) 0.099 0.60 (0.314, 1.154) 0.127

Asthma <0.001 <0.001

Had asthma

Not had asthma -0.59 (-0.711, -0.473) <0.001 0.64 (0.504, 0.816) <0.001

Health insurance 0.373 0.900

49

Covered

Uninsured 0.07 (-0.087, 0.233) 0.373 1.02 (0.760, 1.367) 0.899

Housing condition <0.001 0.638

Own

Rent 0.28 (0.164, 0.390) <0.001 1.12 (0.863, 1.451) 0.398

Other arrangement 0.43 (0.129, 0.732) 0.005 0.96 (0.550, 1.676) 0.887

50

As for the K6 score, no significant difference is identified between construction

workers and non-construction workers (p=0.414) with all covariates adjusted.

The construction workers have a mean K6 score 0.07-point lower than the non-

construction workers (95%CI: -0.251, 0.103).

As for the covariate age, those among 25-44 or 45-65 tend to have a significantly

lower K6 score (p<0.001). People from 25-44 have a mean K6 score 0.45-point

lower than those from 18-24 (95%CI: -0.681,-0.216) and people from 45-65 have

a mean K6 score 0.98-point lower than those from 18-24 (95%CI: -1.211, -

0.745).

As for gender, female tend to have a significantly higher K6 score (p<0.001),

which is 0.33-point higher than that among male (95%CI: 0.252, 0.404). As for

the ethic group, the non- Hispanic tend to have a significantly higher K6 score

(p=0.002), which is 0.33-point higher that that among the Hispanic (95%CI:

0.126, 0.543).

As for the annual family income, the mean K6 score in the group of 15k-20k or

higher is significantly lower than that in the group with annual family income less

than 10k (p<0.05). The difference in the mean K6 score in these groups is 0.68-

point (95%CI: -1.258, -0.107), 1.07-point (95%CI: -1.613, -0.530), 1.34-point

(95%CI: -1.865,-0.816), 1.48-point (95%CI: -1.994. -0.961), 1.70-point (95%CI:

-2.208,-1.183) and 1.72-point (95%CI: -2.234,-1.211) lower than the lowest

category. As for the housing status, those who rent their house tend to have a

significantly higher K6 score (p<0.001), which is estimated to be 0.28-point

higher than those who own their house (95% CI: 0.164, 0.390).

As for smoking status, current non-smokers tend to have a significantly lower K6

score compared with the current smokers (p<0.001). It is estimated that the mean

K6 score of the former group is 0.65-point lower than that among the latter group

(95% CI: -0.790,-0.511).

51

For other covariates, they are also significantly associated with the mean K6

score, except the covariates race, physical activity, had CHD, had stroke and

health insurance coverage.

As for the odds of SMI, with all covariates adjusted, no significant difference is

identified between construction workers and non-construction workers (p=0.672).

The odds among construction workers is 89% of the odds among non-

construction workers (95%CI: 0.532, 1.502).

As for the covariate age, only those among 45-65 tend to have a significantly

lower odds of SMI (p=0.008). The odds among people from 45-65 is 58% of the

odds among those from 18-24 (95%CI: 0.388, 0.867).

As for gender, female tend to have a significantly higher odds of having SMI

(p<0.001), which is 1.48 times the odds among male (95%CI: 1.183, 1.862).

However, as for the ethic group, the non-Hispanic do not have a significantly

higher odds of having SMI (p=0.219) compared with the Hispanic.

As for the annual family income, the odds of SMI in the group of 25k-35k or

higher is significantly lower than that in the group of less than 10k. The odds of

SMI in these groups is 48% (95%CI: 0.284, 0.806), 45% (95%CI: 0.269, 0.762),

30% (95%CI: 0.171, 0.522) and 25% (95%CI: 0.141, 0.433) of the odds in the

lowest category. As for the housing status, no significant difference is found in

the odds of SMI among those who own their house and those who rent their

house (p=0.638).

As for smoking status, current non-smokers tend to have a significantly lower

odds of SMI compared with the current smokers (p<0.001). It is estimated that

odds in the former group is 60% of the odds in the latter group (95% CI: 0.466,

0.763).

For other covariates, they are also significantly associated with SMI, except the

covariates race, physical activity, had CHD, had stroke and health insurance

coverage.

52

5 Reflections on the Existing Data Sets

At the national level, permanent construction workers tend to have a significantly lower

mean K6 score compared with non-construction workers. Temporary construction workers

are found to have a higher mean K6 score than the permanent construction workers, though

the difference is not significant at the 0.05 significant level. In terms of the odds of SMI,

permanent construction workers are found to have significantly lower odds of SMI compared

with non-construction workers. Temporary construction workers are found to have a higher

odds of SMI than the permanent construction workers, though the difference is not

significant at the 0.05 significant level, either.

At the state level, the information on the status of employment is unavailable, thus the

comparisons of the mean K6 score and odds of SMI are conducted only between

construction workers and non-construction workers. The results suggest that construction

workers have a slightly lower mean K6 score and a slightly lower odds of SMI, and both of

them are insignificant. Based on the results from the national level analysis, it is plausible

that temporary construction workers may have a higher K6 score and odds of SMI in

Washington State, too. Thus it would be worthwhile to collect the temporary employment

information at the state level to have a better understanding of this population.

One potential explanation to the significant, lower mean K6 score and odds of SMI found

only at the national level between non- and permanent construction workers could be the

discrepancy in terms of the occupation categorization. Although both MEPS and BRFSS-

WA categorize the respondents’ occupation based on the Standard Occupational

Classification (SOC), they use different strategies to present the information. With MEPS,

people’s occupation information is initially coded at the 4-digit level and it is condensed into

broader groups for the protection of confidentiality. The same situation occurs with BRFSS-

WA where the detailed occupation groups are aggregated by the Washington State

Department of Labor and Industries. Thus at the national level, construction, extraction and

maintenance are considered as one category while at the state level, construction and

extraction are combined as one category but maintenance is in another category along with

installation and repair. Based on the analysis of BRFSS-WA, combing people from

maintenance with those from construction and extraction at the national level may

considerably lower the mean K6 score and odds of SMI for this occupation category.

Because at the state level, the mean K6 score of construction and extraction workers is 2.80

53

but 2.27 for the installation, repair and maintenance workers. Similarly, the prevalence of

SMI of construction and extraction workers is 1.7% but 0.9% for the installation, repair and

maintenance workers. It would be beneficial if both data sources use the same strategy to

present the information, or make the data available at a more detailed level. In addition, one

drawback of condensing the occupation information is losing the variation among different

trades under the same occupation category and obscuring the profile of the occupation as a

whole. For instance, the work condition of construction and building inspectors are very

different from that of sheet metal workers, thus it is reasonable to speculate variance of the

mental health performance between these two trades. It would be beneficial to have the

detailed occupation information available in case when analysis done at the trade level is

warranted.

Apart from the difference in terms of occupation categorization, the discrepancy between

these two data sources also exists when it comes to variable nomenclature. For instance, both

MEPS and BRFSS-WA collect information on the K6 score, however, the score is named as

the Kessler Index by the former and as the Serious Mental Illness Index score by the latter. It

would be beneficial if such variables are named in a way that is widely agreed in the field,

which can ultimately increase the efficiency of data access and integration.

The discontinuity in the data collection is also found in the MEPS and BRFSS-WA. For

instance, in MEPS, the information of the highest degree obtained by the respondents is

collected from 2008 to 2015 except for 2013 and 2014. In BRFSS, the information of

housing types the respondents live in is collected from 2008 to 2016, except for 2011. It

would be helpful to have continuous information available for trend analysis.

Besides, in order to increase the utilization of data sources, it is crucial to step outside the

box by looking into non-typical data sources. For instance, information on the temporary

employment status is collected in MEPS whose major interest is on medical expenditures.

Information on occupational and housing status is collected in BRFSS-WA whose focus is

on people’s health changes. Information of such secondary nature can easily be overlooked

because it sits quietly along with a variety of other information and scatters across different

data sources.

54

The knowledge to leverage this type of information hidden in non-typical data sources can

be potentially very helpful to the research community at large, if the knowledge can be

somehow documented, accumulated and disseminated. Ideally, mapping out the different

data sources and providing ways for data integration will enable the scholars and

stakeholders to better understand the social, economic, or even environmental factors

impacting the construction workforce.

55

6 Conclusion

This research looks into the influence of temporary employment on the mental health of

construction workers under the emerging U.S. social-economic environment. Data from

MEPS and BREFF-WA were extracted and analyzed for this research purpose, with

potential covariates such as income, housing, and lifestyle behaviors being adjusted.

Regardless of the employment status, the general construction workers’ profile indicates a

disadvantaged population compared with the non-construction workers. Although being a

construction worker is not found to be associated with a significantly worse mental health

performance at both the national and the state level, a construction worker does tend to have

exposure to the industry characteristics which make him/her more prone to the negative

outcome. For all the covariates such as age, highest education degree obtained, annual family

income, perceived general health and smoking status which are identified with a significant

negative association with the mental health performance, construction workers are always in

worse conditions compared with non-construction workers.

Furthermore, temporary employment status is found to be associated with worse mental

health performance within the construction industry. Temporary construction workforce

tends to have higher mean K6 score and higher odds of SMI when compared with permanent

construction workforce at the national level. Although the differences are not statistically

significant, it might be due to aggregating maintenance workers with construction and

extraction workers. With regard to the negative factors and when compared with permanent

construction workers, temporary construction workforce consists of even younger workers,

with lower education degree obtained, lower family income, poorer self-rated overall health

and higher current smoking rate. Although there is no information available on the

temporary employment in the BRFSS-WA data set, it is reasonable to speculate a worse

mental health performance in temporary construction workers in the Washington State.

However, due to the lack of data resource, it is still unclear what roles other typical

temporary employment features (e.g. higher job insecurity or exposure to hazardous work

conditions) play in undermining the mental health of this population. Moreover, there is no

unitary method that can address the issue at once, since the characteristics identified are

inherent to this population and require the efforts from all the stakeholders.

56

In order to promote the general wellbeing of the construction workers, more research efforts

are required to raise the awareness and strengthen the understanding of the mentally

demanding nature of the construction occupation. There has been insufficient research done

to address the mental health aspect of the construction industry and attention has been

dominantly focused on physical health. It is worthwhile for future studies to look into how

different trades are possibly associated with better or worse mental health performance, how

the construction industry specific characteristics like physical demands, long working hours,

work-life conflicts etc., reshape the mental health aspect of the workforce, and how the

mental health interacts with other major outcomes of interest such as productivity, injury

rate, safety climate etc. before proper interventions could be devised to promote the mental

health of the construction workers.

Meanwhile, it is also unacceptable if the flexibility guaranteed by the temporary employment

is taken for granted. The variance in the labor force demand is one typical characteristic that

distinguishes the construction industry from the other. It is beneficial for all stakeholders to

address the issues coming along since it is an indispensable element in the overall workforce

with which proactive management is better than ignorance. It is worthwhile for future studies

to explore whether the negative impact of temporary employment is further escalated by the

construction industry characteristics and how these factors are associated with other

outcomes of interest.

Last but not the least, more effort is also required in the data collection process since it is the

fundamental step toward valid and reliable analysis and interpretation. The incompleteness,

inconsistency and discontinuity undermine the chances of such analysis. It is also going to

improve the work efficiency of the researchers if the data set can be aggregated by

standardized coding and documentation without unnecessary repetition, which will optimize

the utilization of the data resource in turn.

57

7 Bibliography

Arndt, V., D. Rothenbacher, U. Daniel, B. Zschenderlein, S. Schuberth, and H. Brenner.

"Construction work and risk of occupational disability: a ten year follow up of 14 474 male

workers." Occupational and environmental medicine 62, no. 8 (2005): 559-566.

Barbeau, Elizabeth, M. Li, Yi Calderon, Patricia Hartman, Cathy Quinn, Margaret

Markkanen, Pia Roelofs, Cora Frazier, and Lindsay Levenstein. "Results of a Union-based

Smoking Cessation Intervention for Apprentice Iron Workers (United States)." Cancer

Causes & Control 17, no. 1 (2006): 53-61.

Bodner, Todd, Mariah Kraner, Brittany Bradford, Leslie Hammer, and Donald Truxillo.

"Safety, health, and well-being of municipal utility and construction workers." Journal of

occupational and environmental medicine 56, no. 7 (2014): 771-778.

Borstad, John D., Blake Buetow, Emily Deppe, Jonas Kyllonen, Marie Liekhus, Cort J.

Cieminski, and Paula M. Ludewig. "A Longitudinal Analysis of the Effects of a Preventive

Exercise Programme on the Factors That Predict Shoulder Pain in Construction

Apprentices." Ergonomics 52, no. 2 (2009): 232-44

Boschman, Van Der Molen, Sluiter, and Frings-Dresen. "Psychosocial Work Environment

and Mental Health among Construction Workers." Applied Ergonomics 44, no. 5 (2013):

748-55.

Caban, Alberto J., David J. Lee, Lora E. Fleming, Orlando Gómez-Marín, William LeBlanc,

and Terry Pitman. "Obesity in US workers: The national health interview survey, 1986 to

2002." American Journal of Public Health 95, no. 9 (2005): 1614-1622.

Coffeng, Jennifer K., Cécile RL Boot, Saskia FA Duijts, Jos WR Twisk, Willem van

Mechelen, and Ingrid JM Hendriksen. "Effectiveness of a worksite social & physical

environment intervention on need for recovery, physical activity and relaxation; results of a

randomized controlled trial." PloS one 9, no. 12 (2014): e114860.

Dong, Xiuwen, Wang Xuanwen, and Katz Rebecca. "The construction chart book." USA:

CPWR-The Center for Construction Research and Training (2018).

58

Dong XS, Wang X, Largay JA. “Second quarter data report: Temporary workers in the

construction industry.” USA: CPWR-The Center for Construction Research and Training

(2015).

Eriksen, H. R., C. Ihlebaek, A. Mikkelsen, H. Grønningsæter, G. M. Sandal, and H. Ursin.

"Improving subjective health at the worksite: a randomized controlled trial of stress

management training, physical exercise and an integrated health programme." Occupational

Medicine 52, no. 7 (2002): 383-391.

Fan, Z. J., Bonauto, D. K., Foley, M. P., Anderson, N. J., Yragui, N. L., & Silverstein, B. A.

(2012). Occupation and the prevalence of current depression and frequent mental distress,

WA BRFSS 2006 and 2008. American Journal of Industrial Medicine, 55(10), 893-903.

Hammer, Leslie B., Truxillo, Donald M., Bodner, Todd, Rineer, Jennifer, Pytlovany, Amy

C., and Richman, Amy. "Effects of a Workplace Intervention Targeting Psychosocial Risk

Factors on Safety and Health Outcomes." BioMed Research International, 2015, BioMed

Research International, Annual, 2015.

Kessler, Ronald C., Gavin Andrews, Lisa J. Colpe, Eva Hiripi, Daniel K. Mroczek, S-LT

Normand, Ellen E. Walters, and Alan M. Zaslavsky. "Short screening scales to monitor

population prevalences and trends in non-specific psychological distress." Psychological

medicine 32, no. 6 (2002): 959-976.

Kessler, Ronald C., Patricia A. Berglund, Shanyang Zhao, Philip J. Leaf, Anthony C.

Kouzis, Martha L. Bruce, Robert M. Friedman et al. "The 12-month prevalence and

correlates of serious mental illness (SMI)." Mental Health, United States 5600 (1996).

Maes, Stan, Chris Verhoeven, France Kittel, and Hetty Scholten. "Effects of a Dutch work-

site wellness-health program: the Brabantia Project." American journal of public health 88,

no. 7 (1998): 1037-1041.

Olson, Ryan, Robert R. Wright, Diane L. Elliot, Jennifer A. Hess, Sharon Thompson, Annie

Buckmaster, Kristy Luther, and Brad Wipfli. "The COMPASS pilot study: a Total Worker

59

Health™ intervention for home care workers." Journal of occupational and environmental

medicine 57, no. 4 (2015): 406-416.

Palumbo, Mary Val, Ge Wu, Hollie Shaner-McRae, Betty Rambur, and Barbara McIntosh.

"Tai Chi for older nurses: a workplace wellness pilot study." Applied Nursing Research 25,

no. 1 (2012): 54-59.

Roberts, Hodgson, and Dolan. ""It's Driving Her Mad": Gender Differences in the Effects of

Commuting on Psychological Health." Journal of Health Economics 30, no. 5 (2011): 1064-

076.

Silla, Inmaculada, Francisco J Gracia, and José María Peiró. "Job Insecurity and Health-

Related Outcomes among Different Types of Temporary Workers." Economic and Industrial

Democracy 26, no. 1 (2005): 89-117.

Sorensen, Glorian, Elizabeth Barbeau, M. Stoddard, Anne Hunt, Mary Goldman, Kay Smith,

Roberta Brennan, and Ann Wallace. "Tools for Health: The Efficacy of a Tailored

Intervention Targeted for Construction Laborers." Cancer Causes & Control 18, no. 1

(2007): 51-59.

Szumilas, Magdalena. "Explaining odds ratios." Journal of the Canadian academy of child

and adolescent psychiatry 19, no. 3 (2010): 227.

Talton, Jon. "Washington State Saw the Hottest GDP Growth in U.S. Last Year." The Seattle

Times, May 17, 2017. Accessed June 4, 2018.

https://www.seattletimes.com/business/economy/state-saw-the-hottest-gdp-growth-in-u-s-

last-year/.

Torres, Blanca. "Washington State Ranks No. 1 for Combined Job and Wage Growth." The

Seattle Times, February 16, 2016. Accessed June 4, 2018.

https://www.seattletimes.com/business/economy/employment-and-wage-growth-in-

washington-outpacing-other-states/.

60

U.S. Department of Health & Human Services, " Health-Related Quality of Life: Well-Being

Concepts", Center for Disease and Control (CDC).

https://www.cdc.gov/hrqol/wellbeing.htm#one (accessed April 1, 2018)

U.S. Department of Health & Human Services, “2014 Poverty Guidelines”, Office of the

Assistant Secretary for Planning and Evaluation (ASPE). https://aspe.hhs.gov/2014-poverty-

guidelines (accessed March 27, 2018)

U.S. Department of Health & Human Services, “Medical Expenditure Panel Survey”, the

Agency for Healthcare Research and Quality (AHRQ) .

https://meps.ahrq.gov/mepsweb/about_meps/survey_back.jsp (accessed March 27, 2018)

U.S. Department of Health & Human Services, “Total Worker Health”, the National Institute

for Occupational Safety and Health (NIOSH).

https://www.cdc.gov/niosh/twh/totalhealth.html (accessed March 27, 2018)

U.S. Department of Health & Human Services, “Stress at Work”, the National Institute for

Occupational Safety and Health (NIOSH). https://www.cdc.gov/niosh/docs/99-

101/default.html (accessed June 1, 2018)

U.S. Department of Labor, “Occupational Employment Statistics”, the Bureau of Labor

Statistics. https://www.bls.gov/oes/oes_emp.htm (accessed April 1, 2018)

U.S. Department of Health & Human Services, “What is mental health”, the

MentalHealth.gov. https://www.mentalhealth.gov/basics/what-is-mental-health (accessed

June 1, 2018)

Virtanen, Marianna, Kivimki, Mika, Joensuu, Matti, Virtanen, Pekka, Elovainio, Marko, and

Vahtera, Jussi. "Temporary Employment and Health: A Review." International Journal of

Epidemiology 34, no. 3 (2005): 610-22.

Wang Xuanwen, Katz Rebecca, and Dong, Xiuwen. “Fourth quarter data report: Health,

Healthcare, and Medical Expenditures among Construction Workers." USA: CPWR-The

Center for Construction Research and Training (2016).

61

Washington State Department of Health (WSDOH), “Behavioral Risk Factor Surveillance

System”.

https://www.doh.wa.gov/DataandStatisticalReports/DataSystems/BehavioralRiskFactorSurv

eillanceSystemBRFSS-WA (accessed March 27, 2018)

62

8 Appendix

Table 11 Descriptive Summary of All the Variables (At national level, 2008)

Non-

construction %

Permanent

construction %

temporary

construction %

Age

28-24 1618 13% 128 11% 26 16%

25-44 5976 48% 642 54% 91 58%

45-65 4971 40% 413 35% 41 26%

Gender

Male 5853 47% 1151 97% 154 97%

Female 6712 53% 32 3% 4 3%

Race

White 8908 71% 1010 85% 138 87%

Black 2314 18% 123 10% 10 6%

American Indian/Alaska

Native 128 1% 8 1% 2 1%

Asian 946 8% 26 2% 2 1%

Native Hawaiian/Pacific

Islander 51 0% 1 0% 1 1%

Multiple races reported 218 2% 15 1% 5 3%

Ethic group

Hispanic 3125 25% 431 36% 109 69%

Not Hispanic 9440 75% 752 64% 49 31%

Marriage status

Married 7089 56% 749 63% 75 47%

Not married 5475 44% 434 37% 83 53%

Highest degree

No Degree 1932 16% 329 28% 85 54%

GED 482 4% 65 6% 7 4%

High School Diploma 5476 44% 622 53% 52 33%

Bachelor's Degree 2321 19% 61 5% 3 2%

Master's Degree 931 7% 11 1% 3 2%

Doctorate Degree 244 2% 0 0% 0 0%

Other Degree 1075 9% 88 7% 7 4%

Annual personal wage

<25k 5605 45% 482 41% 100 63%

25k-50k 4065 32% 474 40% 47 30%

50k-75k 1670 13% 154 13% 9 6%

75k-100k 683 5% 40 3% 2 1%

>100k 542 4% 33 3% 0 0%

Annual family income

<25k 2209 18% 227 19% 57 36%

63

25k-50k 3328 26% 363 31% 53 34%

50k-75k 2474 20% 259 22% 28 18%

75k-100k 1753 14% 154 13% 12 8%

>100k 2801 22% 180 15% 8 5%

Poverty status

Poor/Negative 1144 9% 128 11% 34 22%

Near poor 573 5% 62 5% 15 9%

Low income 1882 15% 189 16% 29 18%

Middle income 4170 33% 445 38% 61 39%

High income 4796 38% 359 30% 19 12%

Perceived general health

Excellent 3299 26% 296 25% 54 34%

Very good 4459 36% 404 34% 45 28%

Good 3730 30% 370 31% 39 25%

Fair 921 7% 95 8% 19 12%

Poor 147 1% 18 2% 1 1%

Obesity

Under weight 184 2% 7 1% 0 0%

Normal 4031 33% 295 25% 33 22%

Overweight 4310 35% 530 46% 82 55%

Obesity 3701 30% 326 28% 34 23%

Smoke status

Current smoker 2071 19% 308 30% 41 29%

Current non-smoker 9021 81% 731 70% 100 71%

Physical activity

Physically active 7139 57% 720 61% 67 43%

Physically inactive 5281 43% 454 39% 90 57%

CHD

Had CHD 255 2% 29 2% 5 3%

Not had CHD 12304 98% 1153 98% 153 97%

Stroke

Had stroke 153 1% 9 1% 1 1%

Not had stroke 12406 99% 1173 99% 157 99%

Asthma

Had asthma 977 8% 63 5% 10 6%

Not had asthma 11584 92% 1119 95% 148 94%

Health insurance

Covered by any private

insurance 9315 74% 732 62% 30 19%

Covered by public

insurance only 847 7% 48 4% 18 11%

Uninsured 2403 19% 403 34% 110 70%

64

Table 12 Descriptive Summary of All the Variables (At national level, 2012)

Non-

construction %

Permanent

construction %

temporary

construction %

Age

28-24 1802 13% 80 6% 19 12%

25-44 6753 47% 665 53% 89 54%

45-65 5732 40% 510 41% 57 35%

Gender

Male 6701 47% 1219 97% 163 99%

Female 7586 53% 36 3% 2 1%

Race

White 9800 69% 1042 83% 143 87%

Black 2810 20% 140 11% 14 8%

American Indian/Alaska

Native 107 1% 10 1% 2 1%

Asian 1268 9% 42 3% 4 2%

Native Hawaiian/Pacific

Islander 76 1% 4 0% 2 1%

Multiple races reported 226 2% 17 1% 0 0%

Ethic group

Hispanic 4030 28% 504 40% 106 64%

Not Hispanic 10257 72% 751 60% 59 36%

Marriage status

Married 7433 52% 757 60% 71 43%

Not married 6854 48% 498 40% 94 57%

Highest degree

No Degree 995 14% 149 24% 21 30%

GED 230 3% 35 6% 5 7%

High School Diploma 3015 44% 358 57% 36 51%

Bachelor's Degree 1359 20% 40 6% 4 6%

Master's Degree 547 8% 0 0% 1 1%

Doctorate Degree 174 3% 0 0% 0 0%

Other Degree 576 8% 51 8% 3 4%

Annual personal wage

<25k 6222 44% 447 36% 118 72%

25k-50k 4499 31% 518 41% 36 22%

50k-75k 1984 14% 197 16% 6 4%

75k-100k 817 6% 57 5% 2 1%

>100k 765 5% 36 3% 3 2%

Annual family income

<25k 2551 18% 217 17% 74 45%

65

25k-50k 3729 26% 387 31% 39 24%

50k-75k 2826 20% 291 23% 22 13%

75k-100k 1890 13% 164 13% 12 7%

>100k 3291 23% 196 16% 18 11%

Poverty status

Poor/Negative 1374 10% 129 10% 39 24%

Near poor 689 5% 68 5% 22 13%

Low income 2396 17% 219 17% 38 23%

Middle income 4777 33% 464 37% 38 23%

High income 5051 35% 375 30% 28 17%

Perceived general health

Excellent 3914 27% 294 23% 55 33%

Very good 5165 36% 434 35% 47 28%

Good 4031 28% 403 32% 50 30%

Fair 1024 7% 109 9% 12 7%

Poor 146 1% 14 1% 1 1%

Obesity

Under weight 189 1% 7 1% 2 1%

Normal 4686 34% 286 23% 44 27%

Overweight 4901 35% 519 43% 81 50%

Obesity 4206 30% 409 33% 36 22%

Smoke status

Current smoker 2107 16% 284 25% 37 25%

Current non-smoker 10766 84% 854 75% 110 75%

Physical activity

Physically active 7254 51% 749 60% 99 60%

Physically inactive 6855 49% 496 40% 65 40%

CHD

Had CHD 237 2% 28 2% 1 1%

Not had CHD 14044 98% 1227 98% 164 99%

Stroke

Had stroke 141 1% 7 1% 0 0%

Not had stroke 14140 99% 1248 99% 165 100%

Asthma

Had asthma 1137 8% 73 6% 9 5%

Not had asthma 13147 92% 1182 94% 156 95%

Health insurance

Covered by any private

insurance 9969 70% 738 59% 41 25%

Covered by public

insurance only 1116 8% 61 5% 17 10%

Uninsured 3202 22% 456 36% 107 65%

66

Table 13 Descriptive Summary of All the Variables (At national level, 2015)

Non-

construction %

Permanent

construction %

temporary

construction %

Age

28-24 1712 13% 84 8% 30 14%

25-44 6477 47% 521 51% 125 57%

45-65 5490 40% 419 41% 65 30%

Gender

Male 6529 48% 1003 98% 213 97%

Female 7150 52% 21 2% 7 3%

Race

White 9288 68% 856 84% 187 85%

Black 2618 19% 100 10% 22 10%

American Indian/Alaska

Native 107 1% 15 1% 2 1%

Asian 733 5% 20 2% 2 1%

Native Hawaiian/Pacific

Islander 478 3% 13 1% 0 0%

Multiple races reported 455 3% 20 2% 7 3%

Ethic group

Hispanic 4002 29% 463 45% 141 64%

Not Hispanic 9677 71% 561 55% 79 36%

Marriage status

Married 6793 50% 609 59% 103 47%

Not married 6886 50% 415 41% 117 53%

Highest degree

No Degree 950 13% 126 23% 46 44%

GED 268 4% 49 9% 10 10%

High School Diploma 3176 44% 288 53% 38 36%

Bachelor's Degree 1410 19% 26 5% 3 3%

Master's Degree 641 9% 2 0% 0 0%

Doctorate Degree 159 2% 3 1% 0 0%

Other Degree 643 9% 53 10% 8 8%

Annual personal wage

<25k 5770 42% 317 31% 118 54%

25k-50k 4307 31% 441 43% 72 33%

50k-75k 1900 14% 171 17% 26 12%

75k-100k 848 6% 70 7% 3 1%

>100k 854 6% 25 2% 1 0%

Annual family income

<25k 2362 17% 149 15% 64 29%

67

25k-50k 3369 25% 316 31% 74 34%

50k-75k 2563 19% 226 22% 44 20%

75k-100k 1920 14% 167 16% 13 6%

>100k 3465 25% 166 16% 25 11%

Poverty status

Poor/Negative 1290 9% 92 9% 44 20%

Near poor 631 5% 57 6% 12 5%

Low income 2117 15% 159 16% 48 22%

Middle income 4586 34% 412 40% 82 37%

High income 5055 37% 304 30% 34 15%

Perceived general health

Excellent 3642 27% 275 27% 63 29%

Very good 4825 35% 340 33% 55 25%

Good 3965 29% 301 29% 70 32%

Fair 1075 8% 94 9% 27 12%

Poor 132 1% 14 1% 4 2%

Obesity

Under weight 141 1% 11 1% 2 1%

Normal 3278 25% 194 20% 43 21%

Overweight 3632 27% 347 35% 79 38%

Obesity 6292 47% 433 44% 85 41%

Smoke status

Current smoker 1632 14% 201 24% 39 22%

Current non-smoker 9829 86% 643 76% 140 78%

Physical activity

Physically active 7029 52% 627 62% 116 53%

Physically inactive 6422 48% 382 38% 102 47%

CHD

Had CHD 198 1% 19 2% 8 4%

Not had CHD 13455 99% 1004 98% 211 96%

Stroke

Had stroke 165 1% 7 1% 0 0%

Not had stroke 13490 99% 1017 99% 219 100%

Asthma

Had asthma 1145 8% 41 4% 10 5%

Not had asthma 12512 92% 982 96% 209 95%

Health insurance

Covered by any private

insurance 9989 73% 623 61% 75 34%

Covered by public

insurance only 1720 13% 85 8% 34 15%

Uninsured 1970 14% 316 31% 111 50%

68

Table 14 Descriptive Summary of All the Variables (At WA level, 2011)

Non-construction % Construction %

Age

18-24 205 4% 10 5%

25-44 1664 31% 63 32%

45-65 3536 65% 127 64%

Gender

Male 2334 43% 174 87%

Female 3071 57% 26 13%

Race

White 4684 90% 172 90%

Black 76 1% 1 1%

American Indian/Alaska

Native 54 1% 4 2%

Asian 161 3% 4 2%

Pacific Islander 22 0% 0 0%

Other race 115 2% 5 3%

Multiple races reported 112 2% 6 3%

Ethic group

Hispanic 389 7% 16 8%

Not Hispanic 5008 93% 183 92%

Marriage status

Married 3427 64% 126 63%

Not married 1967 36% 74 37%

Highest degree

None or kindergarten 7 0% 0 0%

Grade 1-8 110 2% 4 2%

Grade 9-11 135 2% 11 6%

Grade 12 or GED 989 18% 93 47%

College 1-3 years 1618 30% 60 30%

College 4 years or more 2542 47% 32 16%

Annual family income

<10k 41 1% 1 1%

10k-15k 92 2% 1 1%

15k-20k 157 3% 4 2%

20k-25k 345 7% 17 9%

25k-35k 469 9% 21 11%

35k-50k 729 15% 42 22%

50k-75k 1052 21% 44 23%

>75k 2098 42% 59 31%

Perceived general health

69

Excellent 1222 23% 38 19%

Very good 2100 39% 60 30%

Good 1581 29% 73 37%

Fair 439 8% 25 13%

Poor 58 1% 3 2%

Obesity

Under weight 45 1% 1 1%

Normal 1800 35% 65 33%

Overweight 1866 37% 81 41%

Obesity 1389 27% 49 25%

Smoking status

Current smoker 675 13% 51 26%

Current non-smoker 4704 87% 149 75%

Heavy alcohol

consumption

Yes 365 7% 19 10%

No 4902 93% 175 90%

Physical activity

Physically active 4387 82% 150 75%

Physically inactive 941 18% 50 25%

CHD

Had CHD 83 2% 2 1%

Not had CHD 5304 98% 197 99%

Stroke

Had stroke 51 1% 3 2%

Not had stroke 5349 99% 196 98%

Asthma

Had asthma 697 13% 22 11%

Not had asthma 4685 87% 176 89%

Health insurance

Covered 4752 88% 146 73%

Uninsured 643 12% 53 27%

Housing condition

Own 4201 78% 159 80%

Rent 1011 19% 38 19%

Other arrangement 151 3% 2 1%

70

Table 15 Descriptive Summary of All the Variables (At WA level, 2012)

Non-construction % Construction %

Age

18-24 361 6% 15 6%

25-44 2182 34% 105 40%

45-65 3905 61% 144 55%

Gender

Male 3049 47% 248 94%

Female 3399 53% 16 6%

Race

White 5519 87% 215 83%

Black 109 2% 4 2%

American Indian/Alaska

Native 62 1% 3 1%

Asian 214 3% 4 2%

Pacific Islander 45 1% 1 0%

Other race 257 4% 26 10%

Multiple races reported 133 2% 7 3%

Ethic group

Hispanic 437 7% 34 13%

Not Hispanic 5984 93% 229 87%

Marriage status

Married 3902 61% 149 56%

Not married 2521 39% 115 44%

Highest degree

None or kindergarten 12 0% 0 0%

Grade 1-8 101 2% 8 3%

Grade 9-11 155 2% 21 8%

Grade 12 or GED 1191 18% 102 39%

College 1-3 years 1898 29% 92 35%

College 4 years or more 3085 48% 41 16%

Annual family income

<10k 106 2% 7 3%

10k-15k 123 2% 6 2%

15k-20k 239 4% 16 6%

20k-25k 305 5% 19 8%

25k-35k 518 9% 24 10%

35k-50k 768 13% 45 18%

50k-75k 1181 20% 63 25%

>75k 2721 46% 69 28%

Perceived general health

71

Excellent 1388 22% 42 16%

Very good 2570 40% 85 32%

Good 1884 29% 103 39%

Fair 512 8% 30 11%

Poor 88 1% 4 2%

Obesity

Under weight 74 1% 2 1%

Normal 2146 35% 78 30%

Overweight 2259 37% 105 40%

Obesity 1684 27% 78 30%

Smoking status

Current smoker 901 14% 52 20%

Current non-smoker 5505 86% 209 80%

Heavy alcohol consumption

Yes 445 7% 20 8%

No 5893 93% 234 92%

Physical activity

Physically active 5528 86% 212 80%

Physically inactive 916 14% 52 20%

CHD

Had CHD 125 2% 9 3%

Not had CHD 6306 98% 255 97%

Stroke

Had stroke 68 1% 3 1%

Not had stroke 6374 99% 261 99%

Asthma

Had asthma 845 13% 30 11%

Not had asthma 5581 87% 232 89%

Health insurance

Covered 5542 86% 177 67%

Uninsured 891 14% 87 33%

Housing condition

Own 4812 75% 172 65%

Rent 1430 22% 83 31%

Other arrangement 171 3% 8 3%

Building Type

Mobile home 406 6% 36 14%

Detached one-family house 4970 78% 191 73%

Attached home (duplex/etc) 188 3% 13 5%

Building with 1 - 4 apts/condos 184 3% 8 3%

Building with 5 + apts/condos 558 9% 12 5%

72

Apt/Condo, units unknown 0 0% 0 0%

Other 105 2% 3 1%

73

Table 16 Descriptive Summary of All the Variables (At WA level, 2013)

Non-construction % Construction %

Age

18-24 304 7% 12 7%

25-44 1561 34% 62 36%

45-65 2680 59% 99 57%

Gender

Male 2082 46% 162 94%

Female 2463 54% 11 6%

Race

White 3879 87% 145 84%

Black 99 2% 2 1%

American Indian/Alaska

Native 73 2% 3 2%

Asian 156 3% 3 2%

Pacific Islander 29 1% 0 0%

Other race 148 3% 12 7%

Multiple races reported 95 2% 7 4%

Ethic group

Hispanic 307 7% 18 11%

Not Hispanic 4215 93% 153 89%

Marriage status

Married 2701 60% 93 54%

Not married 1820 40% 79 46%

Highest degree

None or kindergarten 8 0% 0 0%

Grade 1-8 69 2% 6 3%

Grade 9-11 99 2% 10 6%

Grade 12 or GED 841 19% 64 37%

College 1-3 years 1291 28% 68 39%

College 4 years or more 2231 49% 25 14%

Annual family income

<10k 86 2% 2 1%

10k-15k 99 2% 5 3%

15k-20k 168 4% 16 10%

20k-25k 270 6% 8 5%

25k-35k 327 8% 13 8%

35k-50k 539 13% 28 17%

50k-75k 748 18% 42 25%

>75k 1977 47% 52 31%

74

Perceived general health

Excellent 1031 23% 27 16%

Very good 1833 40% 58 34%

Good 1300 29% 66 38%

Fair 319 7% 21 12%

Poor 58 1% 1 1%

Obesity

Under weight 57 1% 1 1%

Normal 1540 36% 54 32%

Overweight 1537 36% 70 41%

Obesity 1183 27% 44 26%

Smoking status

Current smoker 588 13% 43 25%

Current non-smoker 3902 87% 126 75%

Heavy alcohol consumption

Yes 340 8% 28 17%

No 4103 92% 141 83%

Physical activity

Physically active 3704 85% 124 77%

Physically inactive 664 15% 38 23%

CHD

Had CHD 63 1% 6 3%

Not had CHD 4471 99% 166 97%

Stroke

Had stroke 31 1% 1 1%

Not had stroke 4509 99% 171 99%

Asthma

Had asthma 650 14% 12 7%

Not had asthma 3888 86% 161 93%

Health insurance

Covered 3863 85% 116 68%

Uninsured 667 15% 54 32%

Housing condition

Own 3269 72% 115 66%

Rent 1133 25% 54 32%

Other arrangement 115 3% 4 2%

Building Type

Mobile home 289 6% 18 10%

Detached one-family house 3453 76% 131 76%

Attached home (duplex/etc) 164 4% 6 3%

Building with 1 - 4 apts/condos 118 3% 8 5%

75

Building with 5 + apts/condos 456 10% 9 5%

Apt/Condo, units unknown 0 0% 0 0%

Other 41 1% 1 1%

76

Table 17 Descriptive Summary of All the Variables (At WA level, 2014)

Non-construction % Construction %

Age

18-24 240 6% 19 11%

25-44 1255 32% 67 37%

45-65 2397 62% 94 52%

Gender

Male 1737 45% 170 94%

Female 2155 55% 10 6%

Race

White 3318 87% 163 91%

Black 84 2% 1 1%

American Indian/Alaska

Native 53 1% 3 2%

Asian 138 4% 3 2%

Pacific Islander 22 1% 0 0%

Other race 95 3% 8 4%

Multiple races reported 89 2% 2 1%

Ethic group

Hispanic 291 8% 16 9%

Not Hispanic 3580 92% 164 91%

Marriage status

Married 2394 62% 101 57%

Not married 1476 38% 76 43%

Highest degree

None or kindergarten 5 0% 1 1%

Grade 1-8 63 2% 4 2%

Grade 9-11 83 2% 12 7%

Grade 12 or GED 671 17% 77 43%

College 1-3 years 1053 27% 60 33%

College 4 years or more 2015 52% 26 14%

Annual family income

<10k 59 2% 6 4%

10k-15k 80 2% 5 3%

15k-20k 138 4% 5 3%

20k-25k 174 5% 10 6%

25k-35k 269 7% 16 10%

35k-50k 450 12% 19 11%

50k-75k 659 18% 41 25%

>75k 1780 49% 65 39%

Perceived general health

77

Excellent 841 22% 40 22%

Very good 1512 39% 56 31%

Good 1179 30% 67 37%

Fair 318 8% 16 9%

Poor 39 1% 1 1%

Obesity

Under weight 57 2% 1 1%

Normal 1235 34% 52 30%

Overweight 1328 36% 81 46%

Obesity 1048 29% 42 24%

Smoking status

Current smoker 451 12% 37 21%

Current non-smoker 3330 88% 138 79%

Heavy alcohol consumption

Yes 282 8% 22 13%

No 3461 92% 150 87%

Physical activity

Physically active 3390 87% 152 84%

Physically inactive 501 13% 28 16%

CHD

Had CHD 64 2% 1 1%

Not had CHD 3819 98% 178 99%

Stroke

Had stroke 34 1% 6 3%

Not had stroke 3847 99% 174 97%

Asthma

Had asthma 569 15% 20 11%

Not had asthma 3304 85% 159 89%

Health insurance

Covered 3563 92% 144 81%

Uninsured 313 8% 34 19%

Housing condition

Own 2760 72% 122 69%

Rent 953 25% 51 29%

Other arrangement 115 3% 5 3%

Building Type

Mobile home 222 6% 22 12%

Detached one-family house 3011 79% 130 73%

Attached home (duplex/etc) 115 3% 3 2%

Building with 1 - 4 apts/condos 99 3% 4 2%

Building with 5 + apts/condos 350 9% 17 10%

78

Apt/Condo, units unknown 0 0% 0 0%

Other 25 1% 1 1%

79

Table 18 Descriptive Summary of All the Variables (At WA level, 2015)

Non-construction % Construction %

Age

18-24 404 7% 20 7%

25-44 2028 34% 118 38%

45-65 3556 59% 169 55%

Gender

Male 2819 47% 284 93%

Female 3169 53% 23 7%

Race

White 4997 85% 260 88%

Black 128 2% 3 1%

American Indian/Alaska

Native 89 2% 8 3%

Asian 256 4% 3 1%

Pacific Islander 40 1% 1 0%

Other race 206 4% 14 5%

Multiple races reported 135 2% 7 2%

Ethic group

Hispanic 513 9% 33 11%

Not Hispanic 5433 91% 272 89%

Marriage status

Married 3617 61% 169 56%

Not married 2345 39% 135 44%

Highest degree

None or kindergarten 14 0% 0 0%

Grade 1-8 86 1% 7 2%

Grade 9-11 167 3% 16 5%

Grade 12 or GED 1066 18% 136 44%

College 1-3 years 1619 27% 104 34%

College 4 years or more 3029 51% 44 14%

Annual family income

<10k 88 2% 9 3%

10k-15k 113 2% 4 1%

15k-20k 185 3% 12 4%

20k-25k 293 5% 15 6%

25k-35k 394 7% 19 7%

35k-50k 612 11% 42 16%

50k-75k 947 18% 54 20%

>75k 2722 51% 113 42%

Perceived general health

80

Excellent 1324 22% 57 19%

Very good 2443 41% 90 29%

Good 1732 29% 122 40%

Fair 424 7% 36 12%

Poor 65 1% 2 1%

Obesity

Under weight 51 1% 1 0%

Normal 1941 35% 87 29%

Overweight 2002 36% 127 43%

Obesity 1519 28% 81 27%

Smoking status

Current smoker 707 12% 75 25%

Current non-smoker 5103 88% 220 75%

Heavy alcohol consumption

Yes 453 8% 35 12%

No 5279 92% 257 88%

Physical activity

Physically active 4715 84% 226 80%

Physically inactive 881 16% 57 20%

CHD

Had CHD 96 2% 9 3%

Not had CHD 5882 98% 294 97%

Stroke

Had stroke 55 1% 4 1%

Not had stroke 5924 99% 302 99%

Asthma

Had asthma 844 14% 29 10%

Not had asthma 5115 86% 275 90%

Health insurance

Covered 5521 93% 250 83%

Uninsured 447 7% 52 17%

Housing condition

Own 4196 70% 192 63%

Rent 1540 26% 103 34%

Other arrangement 237 4% 11 4%

Building Type

Mobile home 360 6% 27 9%

Detached one-family house 4596 77% 230 76%

Attached home (duplex/etc) 189 3% 18 6%

Building with 1 - 4 apts/condos 149 3% 7 2%

Building with 5 + apts/condos 609 10% 18 6%

81

Apt/Condo, units unknown 20 0% 0 0%

Other 35 1% 2 1%

82

Table 19 Descriptive Summary of All the Variables (At WA level, 2016)

Non-construction % Construction %

Age

18-24 339 6% 20 7%

25-44 1950 36% 127 42%

45-65 3092 57% 158 52%

Gender

Male 2626 49% 286 94%

Female 2755 51% 19 6%

Race

White 4464 85% 244 82%

Black 132 3% 7 2%

American Indian/Alaska

Native 55 1% 6 2%

Asian 244 5% 5 2%

Pacific Islander 27 1% 2 1%

Other race 184 3% 20 7%

Multiple races reported 173 3% 12 4%

Ethic group

Hispanic 422 8% 39 13%

Not Hispanic 4919 92% 263 87%

Marriage status

Married 3195 60% 174 57%

Not married 2161 40% 130 43%

Highest degree

None or kindergarten 9 0% 0 0%

Grade 1-8 87 2% 14 5%

Grade 9-11 157 3% 19 6%

Grade 12 or GED 1053 20% 129 42%

College 1-3 years 1466 27% 98 32%

College 4 years or more 2603 48% 44 14%

Annual family income

<10k 65 1% 7 3%

10k-15k 87 2% 7 3%

15k-20k 160 3% 16 6%

20k-25k 255 5% 11 4%

25k-35k 369 8% 21 8%

35k-50k 598 12% 38 14%

50k-75k 809 17% 56 21%

>75k 2514 52% 114 42%

Perceived general health

83

Excellent 1172 22% 62 20%

Very good 2139 40% 120 39%

Good 1603 30% 92 30%

Fair 404 8% 27 9%

Poor 59 1% 4 1%

Obesity

Under weight 69 1% 0 0%

Normal 1669 34% 83 29%

Overweight 1788 36% 116 40%

Obesity 1414 29% 89 31%

Smoking status

Current smoker 608 12% 68 23%

Current non-smoker 4621 88% 225 77%

Heavy alcohol consumption

Yes 389 8% 25 9%

No 4769 92% 265 91%

Physical activity

Physically active 4670 87% 249 82%

Physically inactive 706 13% 56 18%

CHD 5 0

Had CHD 78 1% 3 1%

Not had CHD 5290 99% 300 99%

Stroke

Had stroke 60 1% 2 1%

Not had stroke 5315 99% 302 99%

Asthma

Had asthma 762 14% 26 9%

Not had asthma 4604 86% 279 91%

Health insurance

Covered 4952 92% 254 84%

Uninsured 406 8% 50 16%

Housing condition

Own 3679 69% 186 62%

Rent 1503 28% 107 35%

Other arrangement 187 3% 9 3%

Building Type

Mobile home 288 6% 22 8%

Detached one-family house 3933 76% 223 77%

Attached home (duplex/etc) 193 4% 13 4%

Building with 1 - 4 apts/condos 162 3% 10 3%

Building with 5 + apts/condos 537 10% 20 7%

84

Apt/Condo, units unknown 13 0% 0 0%

Other 26 1% 1 0%